Multi-dimensional medical data analysis and auxiliary decision making system and method

Abstract

The invention provides a multi-dimensional medical data analysis and auxiliary decision-making system and method. The system comprises a data acquisition module, a preprocessing module, a feature extraction module, a model training module, a multi-dimensional analysis module, a decision-making generation module and a feedback optimization module. The data acquisition module acquires data from multiple sources; the preprocessing module performs data cleaning and standardization; the feature extraction module extracts multi-dimensional features; the model training module trains an adaptive decision model; according to the invention, scientificity, accuracy and reliability of medical decision making can be improved, and powerful support is provided for clinical decision making.

Description

Technical Field

[0001] The present invention relates to the technical field of medical data analysis, and more specifically, the present invention relates to a multi-dimensional medical data analysis and auxiliary decision-making system and method. Background Art

[0002] In the field of medical data analysis and auxiliary decision-making, with the continuous development of medical technology, the scale and complexity of medical data are increasing day by day. How to effectively extract valuable information from massive data and provide a scientific basis for clinical decision-making has become a hot issue in current research. Existing medical data analysis methods usually rely on single-dimensional data, such as electronic medical records or medical images, and lack the ability to comprehensively analyze multi-source heterogeneous data. In addition, traditional data analysis methods have deficiencies in dealing with noisy data, missing values, and data standardization, resulting in limited accuracy and reliability of analysis results. In terms of model training and decision generation, existing technologies often adopt static weight allocation and fixed parameter optimization strategies, which are difficult to adapt to complex and changing clinical scenarios, and lack an effective decision verification mechanism, making it difficult to ensure the reliability and practicality of decisions.

[0003] In the process of implementing the embodiments of the present invention, the inventors found that there are at least the following problems or defects in the prior art: existing medical data analysis systems cannot effectively integrate multi-source heterogeneous data, lack a dynamic weight allocation and optimization mechanism, and the decision generation process lacks multi-stage logical judgment and expert rule verification, making it difficult to meet the high reliability and practicality requirements of clinical decision-making. Summary of the Invention

[0004] The present invention provides a multi-dimensional medical data analysis and auxiliary decision-making system, including: A data acquisition module, configured to obtain structured and unstructured data from multi-source medical devices and databases; A data preprocessing module, configured to perform noise filtering, missing value filling, and standardization processing on the data output by the data acquisition module; A feature extraction module, configured to extract multi-dimensional features from the preprocessed data through a dynamic weight allocation algorithm; A model training module, configured to train an adaptive decision model based on the extracted multi-dimensional features and update the model parameters by using a dynamic weight optimization algorithm; A multi-dimensional analysis module, configured to perform spatial distribution analysis, time series evolution analysis, and correlation analysis on the feature data; A decision generation module, configured to generate clinical decision suggestions according to the multi-dimensional analysis results and verify the decision reliability through multi-stage logical judgment; A feedback optimization module, configured to adjust the model parameters and weight allocation strategy according to the actual application feedback; The feedback optimization module includes: A parameter correction unit that reversely adjusts the penalty term weight of the model loss function according to the decision execution result; A weight decay unit that exponentially decays the feature weights that have not been updated for a long time according to the decay rate and the non-update time interval; A version iteration unit that generates a new model version when the cumulative error exceeds a preset threshold.

[0005] As a further improvement of this application, the data acquisition module includes: A multi-source interface unit for connecting to an electronic medical record system, medical imaging equipment, and wearable sensors; A dynamic cleaning unit for real-time filtering of invalid data based on a preset data quality threshold; A real-time storage unit for establishing a distributed index storage structure according to data types and timestamps.

[0006] As a further improvement of this application, the data preprocessing module includes: A time series alignment unit for interpolating and aligning multi-source asynchronous data according to a unified time reference; An anomaly correction unit for dynamically correcting data that exceeds the range of the mean plus or minus three standard deviations of the current data window. The corrected data is generated according to the relationship between the dynamic adjustment coefficient, the mean, and the standard deviation; A normalization unit for mapping the data to the interval from zero to one.

[0007] As a further improvement of this application, the specific steps of the dynamic weight allocation algorithm include: Calculate the sum of the absolute values of the correlation coefficients between each feature dimension and the clinical indicators, and divide it by the maximum absolute value of the correlation coefficient to obtain the feature importance score; According to the feature importance score, dynamically allocate weights through exponential function normalization; Input the weighted features into the convolutional attention network for fusion.

[0008] As a further improvement of this application, the parameter update process of the dynamic weight optimization algorithm includes: Based on the adaptive learning rate, momentum factor, and regularization coefficient, combined with the historical parameter changes and the gradient of the loss function, dynamically adjust the model parameters.

[0009] As a further improvement of this application, the multi-stage logical judgment includes: An initial judgment stage for matching preset clinical pathway rules; A confidence evaluation stage for calculating the decision confidence through the average difference between the model prediction probability and the historical verification labels; An expert rule verification stage for calling the expert knowledge base for secondary verification when the confidence is lower than the preset threshold.

[0010] As a further improvement of the present application, the spatial distribution analysis includes: Identifying data abnormal aggregation regions through a density clustering algorithm with dynamically adjusted neighborhood radius; Generating a three-dimensional visualization map of the feature distribution density; Establishing a pathological feature transfer matrix between different spatial regions.

[0011] As a further improvement of the present application, the density clustering algorithm includes: Dynamically adjusting the neighborhood radius according to the average distance, distance standard deviation and the number of data points; Merging overlapping clusters using a two-way linking strategy; Judging the probability attribution of boundary points.

[0012] A multi-dimensional medical data analysis and auxiliary decision-making method includes the following steps: Obtaining structured and unstructured data from multi-source medical devices and databases; Performing noise filtering, missing value filling and standardization processing on the obtained structured and unstructured data; Extracting multi-dimensional features from the preprocessed data through a dynamic weight allocation algorithm; Training an adaptive decision model based on the extracted multi-dimensional features, and updating the model parameters using a dynamic weight optimization algorithm; Performing spatial distribution analysis, temporal evolution analysis and correlation analysis on the feature data; Generating clinical decision-making suggestions according to the multi-dimensional analysis results, and verifying the decision reliability through multi-stage logical judgment; Adjusting the model parameters and weight allocation strategy according to the actual application feedback.

[0013] According to the above embodiments of the present invention, it has at least the following beneficial effects: The system can effectively integrate multi-source heterogeneous medical data, accurately extract multi-dimensional features and train an adaptive decision model through dynamic weight allocation and optimization algorithms, improving the accuracy and reliability of medical data analysis. The system can perform spatial distribution, temporal evolution and correlation analysis, providing comprehensive and scientific basis for clinical decision-making, and effectively assisting medical staff to formulate more accurate treatment plans.

[0014] The system can also verify the reliability of the decision through multi-stage logical judgment according to the actual application feedback, and dynamically adjust the model parameters and weight allocation strategy to continuously optimize the system performance. At the same time, the system has efficient data acquisition, preprocessing and storage capabilities, can filter invalid data in real time, fill missing values, perform standardization processing, and establish a distributed index storage structure to ensure the integrity and availability of data, providing strong support for medical data analysis and decision-making. Description of the Drawings

[0015] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features, and advantages of the exemplary embodiments of the present invention will become readily understood. In the drawings, several embodiments of the present invention are shown by way of illustration and not limitation, wherein: Figure 1 FIG. 5 is a schematic structural diagram of a multi-dimensional medical data analysis and auxiliary decision-making system provided by an embodiment of the present invention; Figure 2 FIG. 8 is a schematic flowchart of a multi-dimensional medical data analysis and auxiliary decision-making method provided by an embodiment of the present invention. Detailed Embodiments

[0016] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the present invention, and do not limit the scope of the present invention in any way. On the contrary, these embodiments are provided to make the present invention more thorough and complete, and to convey the scope of the present invention fully to those skilled in the art.

[0017] Those skilled in the art know that the embodiments of the present invention can be implemented as a system, device, equipment, method, or computer program product. Therefore, the present invention can be specifically implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0018] It should be noted that any number of elements in the drawings is for illustration rather than limitation, and any naming is only for distinction and does not have any limiting meaning.

[0019] Embodiment 1: Please refer to Figure 1 , Figure 1 FIG. 5 is a schematic structural diagram of a multi-dimensional medical data analysis and auxiliary decision-making system provided by an embodiment of the present invention. As Figure 1 shown, a multi-dimensional medical data analysis and auxiliary decision-making system 100 includes: A data acquisition module 101, configured to obtain structured and unstructured data from multi-source medical devices and databases; A data preprocessing module 102, configured to perform noise filtering, missing value filling, and normalization processing on the data output by the data acquisition module; A feature extraction module 103, configured to extract multi-dimensional features from the preprocessed data through a dynamic weight allocation algorithm; A model training module 104, configured to train an adaptive decision-making model based on the extracted multi-dimensional features and update the model parameters by using a dynamic weight optimization algorithm; The multi-dimensional analysis module 105 is used to perform spatial distribution analysis, temporal evolution analysis, and correlation analysis on the feature data; The decision-making generation module 106 is used to generate clinical decision-making suggestions based on the multi-dimensional analysis results and verify the reliability of the decisions through multi-stage logical judgment; The feedback optimization module 107 is used to adjust the model parameters and weight allocation strategy according to the actual application feedback.

[0020] It should be noted that the embodiment of the present invention provides a multi-dimensional medical data analysis and auxiliary decision-making system, which mainly includes a data acquisition module, a data preprocessing module, a feature extraction module, a model training module, a multi-dimensional analysis module, a decision-making generation module, and a feedback optimization module. The data acquisition module is used to obtain structured and unstructured data from multi-source medical devices and databases. The multi-source medical devices include electronic medical record systems, medical imaging devices, wearable sensors, etc., and the databases cover various medical-related databases. The data preprocessing module performs noise filtering, missing value filling, and standardization processing on the collected data to ensure the quality and usability of the data. The feature extraction module extracts multi-dimensional features from the preprocessed data through a dynamic weight allocation algorithm, providing a basis for subsequent model training. The model training module trains an adaptive decision-making model based on the extracted multi-dimensional features and updates the model parameters using a dynamic weight optimization algorithm to improve the accuracy and adaptability of the model. The multi-dimensional analysis module performs spatial distribution analysis, temporal evolution analysis, and correlation analysis on the feature data, providing comprehensive data support for decision-making generation. The decision-making generation module generates clinical decision-making suggestions based on the multi-dimensional analysis results and verifies the reliability of the decisions through multi-stage logical judgment, ensuring the scientificity and practicality of the decisions. The feedback optimization module adjusts the model parameters and weight allocation strategy according to the actual application feedback to continuously optimize the system performance.

[0021] Specifically, the multi-source interface unit in the data acquisition module is responsible for connecting various medical devices and databases to ensure stable data transmission. The dynamic cleaning unit filters invalid data in real time based on preset data quality thresholds to ensure data accuracy and integrity. The real-time storage unit establishes a distributed index storage structure according to data types and timestamps for fast data retrieval and management. The time series alignment unit in the data preprocessing module interpolates and aligns multi-source asynchronous data based on a unified time reference to ensure data time consistency. The anomaly correction unit uses a specific formula to correct data deviating from the normal range, where the mean and standard deviation of the current data window are used to determine whether the data is abnormal, and the dynamic adjustment coefficient is adjusted according to the actual situation. The normalization unit maps the data to the [0,1] interval to eliminate the dimensional differences between different features. The dynamic weight allocation algorithm in the feature extraction module calculates the importance scores of each feature dimension, dynamically allocates weights according to the scores, and inputs the weighted features into the convolutional attention network for fusion to extract more representative features. The dynamic weight optimization algorithm in the model training module optimizes the model parameters through a specific parameter update formula, combined with an adaptive learning rate, momentum factor, regularization coefficient, etc., to improve the training effect of the model. The spatial distribution analysis in the multi-dimensional analysis module identifies data abnormal aggregation regions through an improved DBSCAN algorithm and generates three-dimensional visualization maps. At the same time, a pathological feature transfer matrix between different spatial regions is established to provide intuitive data support for clinical decision-making. The multi-stage logical judgment in the decision generation module includes an initial judgment stage, a confidence evaluation stage, and an expert rule verification stage to ensure the reliability of the decision through step-by-step verification. The parameter correction unit in the feedback optimization module reversely adjusts the penalty term weight of the model loss function according to the decision execution result. The weight decay unit decays the feature weights that have not been updated for a long time. The version iteration unit generates a new model version when the cumulative error exceeds the threshold to continuously optimize the system performance.

[0022] Preferably, the multi-source interface unit in the data acquisition module can be connected to medical devices and databases using multiple communication protocols to meet the data transmission requirements of different devices and systems. The preset data quality threshold of the dynamic cleaning unit can be adjusted according to the actual data situation to ensure the accuracy and effectiveness of data filtering. The distributed index storage structure of the real-time storage unit can adopt multiple indexing methods, such as time partition layer, feature coding layer, and fast retrieval layer, etc., to improve the efficiency of data storage and retrieval. The time series alignment unit in the data preprocessing module can adopt multiple interpolation methods, such as linear interpolation, spline interpolation, etc., to meet the time alignment requirements of different data. The dynamic adjustment coefficient of the anomaly correction unit can be adjusted according to the data distribution and actual needs to improve the accuracy of anomaly data correction. The normalization unit can adopt multiple normalization methods, such as min-max normalization, Z-score normalization, etc., to meet the normalization requirements of different data. The dynamic weight allocation algorithm in the feature extraction module can adopt multiple weight allocation strategies.

[0023] In some embodiments, the data acquisition module includes: A multi-source interface unit for connecting to the electronic medical record system, medical imaging equipment, and wearable sensors; A dynamic cleaning unit for real-time filtering of invalid data based on a preset data quality threshold; A real-time storage unit for establishing a distributed index storage structure according to data type and timestamp.

[0024] It should be noted that the data acquisition module plays a crucial role in this system. It is responsible for obtaining structured and unstructured data from multiple medical data sources, providing a basis for subsequent data analysis and decision-making. The multi-source interface unit is a key component of the data acquisition module. It is connected to medical devices and systems such as the electronic medical record system, medical imaging equipment, and wearable sensors through multiple interface protocols to ensure stable data transmission and acquisition. The dynamic cleaning unit monitors and filters invalid data in real time based on a preset data quality threshold to ensure data accuracy and reliability. The real-time storage unit establishes a distributed index storage structure according to data type and timestamp, facilitating rapid data retrieval and management, and providing support for subsequent data preprocessing and analysis.

[0025] In some embodiments, the data acquisition module is configured to pass through a medical device protocol adaptation layer, which integrates DICOM3.0, HL7, and FHIR protocol conversion interfaces to achieve stable docking with CT / MRI imaging equipment, wearable monitoring equipment, and electronic medical record databases. During the data acquisition process, a double-layer filtering mechanism based on a rule engine and LSTM anomaly detection is adopted to ensure the high quality of the acquired data.

[0026] The first - layer filtering uses a Bayesian classifier to evaluate the confidence of unstructured text data. The unstructured text data here mainly includes doctor's handwritten notes, imaging diagnosis reports, etc. The Bayesian classifier calculates the probability that each unstructured text data belongs to valid data based on prior probabilities and sample data, and sets a confidence threshold. Data below this threshold will be preliminarily screened. For example, if the confidence threshold is set to 0.6, when the Bayesian classifier calculates that the confidence of a certain doctor's handwritten note is 0.5, then this note will enter the next step of processing or be directly discarded, depending on the subsequent set rules.

[0027] The second - layer filtering identifies abnormal device data streams through a time - series similarity detection algorithm. This algorithm is based on technologies such as dynamic time warping (DTW). It analyzes the collected device data in time series, compares the time - series pattern of the normal device data. If the similarity of a certain data stream to the normal pattern is lower than the preset similarity threshold, then this data stream is determined to be abnormal. For example, the heart - rate data of a wearable monitoring device should show a certain fluctuation pattern under normal circumstances. When the algorithm detects that the fluctuation pattern of a certain heart - rate data is significantly different from the normal pattern and the similarity is lower than 0.7 (assuming the threshold), it is considered that this heart - rate data stream is abnormal.

[0028] Specifically, the multi - source interface unit needs to have multiple interface protocols to adapt to the data - transmission requirements of different medical devices and systems. For example, electronic medical record systems usually use standard protocols such as HL7 and DICOM for data transmission, while wearable sensors may use wireless communication protocols such as Bluetooth and Wi - Fi. The preset data - quality threshold of the dynamic cleaning unit can be adjusted according to the actual data situation to ensure the accuracy and effectiveness of data filtering. For example, for heart - rate data, a reasonable heart - rate range can be set, and data outside this range will be regarded as invalid data and filtered. The distributed index storage structure of the real - time storage unit can be designed according to the characteristics of the data and query requirements. For example, for time - series data, a time - partition layer can be used for storage to improve the data retrieval efficiency.

[0029] Preferably, the multi - source interface unit can adopt a modular design. According to the interface requirements of different medical devices and systems, the corresponding interface modules can be flexibly configured to improve the compatibility and expandability of the system. The dynamic cleaning unit can combine machine - learning algorithms to evaluate and predict the data quality in real - time, and automatically adjust the data - quality threshold to adapt to the changes of different data sources and data types. The real - time storage unit can adopt distributed database technologies such as Hadoop and Cassandra to achieve high availability and high expandability of data. At the same time, technologies such as data compression and caching can be combined to optimize the data storage and retrieval performance.

[0030] In some embodiments, the data preprocessing module includes: A timing alignment unit for interpolating and aligning multi-source asynchronous data based on a unified time reference; An anomaly correction unit for dynamically correcting data outside the range of the mean of the current data window plus or minus three standard deviations. The corrected data is generated based on the relationship between the dynamic adjustment coefficient, the mean, and the standard deviation; A normalization unit for mapping the data to the interval from zero to one.

[0031] Further, the anomaly correction unit corrects the data deviating from the normal range using the following formula:

[0032] Where, is the corrected data, x is the original data, is the mean of the current data window, is the standard deviation, and k is the dynamic adjustment coefficient; It should be noted that the data preprocessing module plays a crucial role in this system. It is responsible for noise filtering, missing value filling, and standardization processing of the data output by the data acquisition module to ensure the quality and usability of the data. The timing alignment unit interpolates and aligns multi-source asynchronous data based on a unified time reference to ensure the temporal consistency of data from different sources. The anomaly correction unit corrects the data deviating from the normal range using a specific formula to eliminate the impact of abnormal data on the analysis results. The normalization unit maps the data to the [0, 1] interval to eliminate the dimensional differences between different features and provide standardized data for subsequent feature extraction and model training.

[0033] Specifically, the timing alignment unit aligns the data with different timestamps to a unified time reference through an interpolation method. For example, for heart rate data and blood pressure data, if their timestamps are inconsistent, the timing alignment unit can use methods such as linear interpolation or spline interpolation to align them to a time reference of per minute or per second. In the formula of the anomaly correction unit, the mean and standard deviation of the current data window are used to determine whether the data is abnormal, and the dynamic adjustment coefficient is adjusted according to the actual situation. For example, for blood glucose data, if the mean of the current data window is 5.0 mmol / L and the standard deviation is 0.5 mmol / L, when the blood glucose value exceeds 7.5 mmol / L or is lower than 2.5 mmol / L, it will be regarded as abnormal data and corrected. The normalization unit maps the data to the [0, 1] interval. For example, for age data, the age value can be mapped to the [0, 1] interval through the min-max normalization method, that is, age normalization value = (age - minimum age) / (maximum age - minimum age).

[0034] Preferably, the timing alignment unit can adopt various interpolation methods, such as linear interpolation, spline interpolation, nearest neighbor interpolation, etc., to adapt to the time alignment requirements of different data. The dynamic adjustment coefficient of the anomaly correction unit can be adjusted according to the data distribution and actual requirements. For example, for some key indicators, the dynamic adjustment coefficient can be appropriately increased to improve the accuracy of anomaly data correction. The normalization unit can adopt various normalization methods, such as min-max normalization, Z-score normalization, decimal scaling normalization, etc., to meet the normalization requirements of different data. In addition, the data preprocessing module can also include a data cleaning unit for removing duplicate data, abnormal data, and irrelevant data to further improve the data quality.

[0035] In some embodiments, the specific steps of the dynamic weight allocation algorithm include: Step S31: Calculate the sum of the absolute values of the correlation coefficients between each feature dimension and the clinical indicators, and divide it by the maximum absolute value of the correlation coefficient to obtain the feature importance score; Step S32: According to the feature importance score, allocate dynamic weights through exponential function normalization; Step S33: Input the weighted features into the convolutional attention network for fusion.

[0036] Further, the following formula is used in step S31 to obtain the feature importance score:

[0037] where is the correlation coefficient between the i-th feature and the j-th clinical indicator, is the importance score of the i-th feature, and n is the total number of clinical indicators; The following formula is used in step S32 to normalize and allocate dynamic weights:

[0038] where is the dynamic weight of the i-th feature, is the importance score of the k-th feature, and m is the total number of features.

[0039] It should be noted that the dynamic weight allocation algorithm in the feature extraction module is one of the key technologies of this system. It calculates the importance scores of each feature dimension, dynamically allocates weights according to the scores, so as to achieve the effective extraction and fusion of multi-dimensional features. Among them, the importance score of the feature dimension is determined by calculating the correlation coefficient between the i-th feature and the j-th clinical index. The larger the absolute value of the correlation coefficient, the stronger the correlation between the feature and the clinical index, and the higher its importance score. The dynamic weight is obtained after normalizing the importance scores of each feature, and is used to measure the relative importance of each feature in the model. By inputting the weighted features into the convolutional attention network for fusion, more representative and discriminative features can be further extracted, providing strong support for subsequent model training and decision-making generation.

[0040] Specifically, when calculating the importance scores of each feature dimension, it is necessary to traverse the correlation coefficients of all features and all clinical indexes, sum the absolute values of the correlation coefficients, and then divide by the total number of clinical indexes to obtain the importance score of each feature. In terms of specific parameter settings, the total number of clinical indexes should be determined according to the actual clinical needs and data conditions. For example, it can include multiple indexes such as heart rate, blood pressure, and blood sugar. When determining the dynamic weight, the importance scores of each feature are normalized, that is, the importance score of each feature is divided by the sum of the importance scores of all features to obtain the dynamic weight of each feature. These weights reflect the relative importance of each feature in the model. The larger the weight value, the greater the contribution of the feature to the model. In addition, the convolutional attention network is a deep learning model that combines convolutional neural network and attention mechanism, which can effectively extract local and global information of features and further improve the expression ability of features.

[0041] Preferably, when calculating the correlation coefficient between the feature and the clinical index, various correlation measurement methods can be used, such as Pearson correlation coefficient, Spearman rank correlation coefficient, etc., to adapt to different data types and distribution characteristics. For data with obvious time series characteristics, dynamic time warping distance and other measurement methods can also be considered to better capture the temporal correlation between the feature and the clinical index. When inputting the weighted features into the convolutional attention network for fusion, various feature fusion strategies can be used, such as element-wise multiplication, element-wise addition, splicing, etc., to achieve effective interaction and fusion between different features. In addition, transfer learning technology can be combined to apply the pre-trained convolutional attention network model to this system to improve the training efficiency and generalization ability of the model.

[0042] In some embodiments, the parameter update process of the dynamic weight optimization algorithm includes: Based on an adaptive learning rate, momentum factor, and regularization coefficient, combining historical parameter changes with the gradient of the loss function, dynamically adjust the model parameters. The parameter update formula of the dynamic weight optimization algorithm is as follows:

[0043] where, is the model parameter at the (t + 1)-th iteration, is the model parameter at the t-th iteration, is the model parameter at the (t - 1)-th iteration, t is the number of iterations, is the adaptive learning rate, is the momentum factor, is the regularization coefficient, L is the loss function, is the gradient of the loss function at the i-th iteration; It should be noted that the dynamic weight optimization algorithm in the model training module is a key technology to improve the model performance. Its core lies in continuously iteratively updating the model parameters to minimize the loss function and improve the generalization ability of the model. The parameter update formula of this algorithm comprehensively considers factors such as the parameters of the previous iteration, the gradient of the loss function, the adaptive learning rate, the momentum factor, and the regularization coefficient. Among them, the number of iterations is used to record the current iteration round. The adaptive learning rate is dynamically adjusted according to the iterative process to ensure that the model can converge quickly in the initial stage of training and then make fine adjustments in the later stage. The momentum factor is used to accelerate the convergence speed in the gradient descent process, and the regularization coefficient helps to prevent the model from overfitting. The loss function is used to measure the difference between the model prediction value and the actual value. Through the reasonable setting and optimization of the above parameters, the model can more accurately learn the complex patterns in the data, thereby improving the reliability of clinical decisions.

[0044] Specifically, each parameter in the parameter update formula needs to be carefully set according to the actual data and model training situation. For example, the adaptive learning rate can be dynamically adjusted according to the change of the loss function. The initial value is usually set to 0.01 or 0.001 and gradually decreased according to the set decay strategy during the training process. The momentum factor generally takes values between 0.5 and 0.9, which is used to accelerate convergence and reduce oscillations. The regularization coefficient is adjusted according to the complexity of the data and the scale of the model, usually selected between 0.0001 and 0.1. The loss function can adopt various forms such as mean squared error, cross entropy, etc. The specific selection depends on the task type of the model and the data characteristics. In addition, the setting of the number of iterations needs to balance the training time and performance of the model. Usually, the best number of iterations is determined by the performance on the validation set to avoid overfitting or underfitting.

[0045] Preferably, when implementing the dynamic weight optimization algorithm, various strategies can be adopted to further improve the training effect of the model. For example, second-order derivative information such as RMSprop or Adam optimizer can be introduced to more effectively adjust the learning rate and improve the model's navigation ability in complex loss landscapes. In addition, an early stopping strategy can be adopted. When the performance on the validation set does not improve significantly for several consecutive iterations, the training is terminated early to save computing resources and prevent overfitting. Transfer learning can also be combined, using the parameters of the pre-trained model as initial parameters and fine-tuning to adapt to a specific medical dataset, thereby accelerating the model's convergence and improving its generalization performance.

[0046] In some embodiments, the multi-stage logical judgment includes: Initial judgment stage: Matching the preset clinical pathway rules; Confidence evaluation stage: Calculating the decision confidence through the average difference between the model prediction probability and the historical verification label; Expert rule verification stage: When the confidence is lower than the preset threshold, call the expert knowledge base for secondary verification.

[0047] Furthermore, in the confidence evaluation stage: The decision confidence is calculated using the following formula:

[0048] where pk is the prediction probability of the model for the k-th sample, yk is the historical verification label of the k-th sample, and k is the total number of historical verification samples; Expert rule verification stage: When 0.85, call the expert knowledge base for secondary verification.

[0049] It should be noted that the multi-stage logical judgment in the decision generation module is an important link to ensure the reliability of clinical decision-making recommendations. This process includes an initial judgment stage, a confidence evaluation stage, and an expert rule verification stage. In the initial judgment stage, the system matches the preset clinical pathway rules according to the decision category output by the model to ensure the consistency of the decision with the clinical standard process. The confidence evaluation stage quantifies the confidence level of the model in the decision by calculating the decision confidence, and the calculation of the confidence is based on the difference between the prediction probability of the model for the sample and the historical verification label. When the confidence is lower than the set threshold (such as 0.85), the system will enter the expert rule verification stage and call the expert knowledge base for secondary verification to ensure the accuracy and reliability of the decision.

[0050] Specifically, the clinical pathway rules in the initial judgment stage are a series of decision rules preset according to clinical guidelines and expert experience, which are used to guide the decision-making categories output by the model. These rules may include the standard processes for disease diagnosis, the selection criteria for treatment plans, etc. In the calculation formula of the confidence evaluation stage, the predicted probability reflects the decision-making confidence of the model for each sample, the historical verification label is the accurate label verified based on past data, and the total number of samples is used to calculate the average confidence. The trigger condition for the expert rule verification stage is that the confidence is lower than 0.85, and this threshold can be adjusted according to actual application requirements and model performance. The expert knowledge base contains rich clinical experience and rules, which are used to provide additional decision-making support when the confidence is insufficient.

[0051] In some embodiments, the decision generation module implements a three-level verification mechanism: First-stage verification: Check the logical consistency between the decision recommendation and the clinical data through a causal inference engine. The causal inference engine constructs a causal relationship model between clinical data based on technologies such as Bayesian networks. Input the decision recommendation into this model and check whether the decision conforms to the causal logic between clinical data. For example, if the decision recommendation is to use a certain drug for a patient, but the patient's allergy history shows an allergy to this drug, the causal inference engine will determine that the decision recommendation is logically inconsistent with the clinical data.

[0052] Second-stage verification: Conduct an evidence-based medicine compliance assessment based on a knowledge graph (including structured data from UpToDate and NCCN guidelines). The knowledge graph integrates authoritative medical guidelines and clinical research results, and matches and compares the decision recommendation with the information in the knowledge graph. For example, for the treatment decision of a certain disease, check whether the decision conforms to the treatment plans recommended in UpToDate and NCCN guidelines, calculate the similarity between the decision recommendation and the guideline content, and if the similarity is lower than a preset threshold (such as 0.8), it is considered that the evidence-based medicine compliance of the decision recommendation is insufficient.

[0053] Third-stage verification: Invoke the expert review interface and adopt a double-blind voting mechanism to manually confirm high-risk decisions. When a decision is determined to be high-risk (such as involving major surgeries, using high-risk drugs, etc.), start the expert review process. Invite multiple medical experts to evaluate the decision. The experts vote without knowing the specific source of the decision and the opinions of other experts. If the majority of experts (such as more than 2 / 3 of the experts) approve the decision, the decision passes the verification; otherwise, the decision needs to be further adjusted or re-evaluated.

[0054] Preferably, when implementing multi-stage logical judgment, the parameters and processes of each stage can be optimized. For example, in the initial judgment stage, the clinical pathway rules can be updated regularly to incorporate the latest clinical research results and treatment guidelines. In the confidence assessment stage, in addition to calculating the overall confidence, the confidence of each sample can be evaluated individually to identify potential misjudged samples. Furthermore, more expert rule verification mechanisms, such as machine learning-based anomaly detection models, can be introduced to further improve the reliability of the decision-making. When the confidence is below the threshold, the system can also automatically record the relevant sample and decision information for subsequent analysis and model optimization.

[0055] In some embodiments, the feedback optimization module performs the following operations: Parameter correction unit: reversely adjusts the penalty term weight of the model loss function according to the decision execution result; Weight decay unit: exponentially decays the feature weights that have not been updated for a long time according to the decay rate and the non-update time interval; Version iteration unit: generates a new model version when the cumulative error exceeds a preset threshold.

[0056] Furthermore, the weight decay unit decays the feature weights that have not been updated for a long time using the following formula:

[0057] where, is the weight of the i-th feature after update, is the weight of the i-th feature before update, is the decay rate, and t is the non-update time interval.

[0058] It should be noted that the feedback optimization module in this system is responsible for adjusting the model parameters and weight allocation strategy according to the actual application feedback to continuously optimize the system performance. This module includes a parameter correction unit, a weight decay unit, and a version iteration unit. The parameter correction unit reversely adjusts the penalty term weight of the model loss function according to the decision execution result to correct the bias of the model in the decision-making process. The weight decay unit decays the feature weights that have not been updated for a long time to reduce the influence of these features on the model and prevent overfitting. The version iteration unit generates a new model version when the cumulative error exceeds a preset threshold to ensure the continuous optimization and performance improvement of the system.

[0059] Specifically, the parameter correction unit adjusts the penalty term weight in the model loss function by analyzing the difference between the decision execution result and the expected result. For example, if a certain feature causes a large error in the decision-making process, the penalty term weight of this feature will be increased to reduce its negative impact on the model. The weight decay unit uses a specific decay formula to adjust the weights of features that have not been updated for a long time. The decay rate and the non-update time interval are the key parameters of this unit and can be set according to the actual application requirements. For example, for features that have not been updated within a month, the decay rate can be set to 0.1, that is, the weight decays by 10% per month. The version iteration unit triggers the generation of a new model version by monitoring the cumulative error of the model. The preset threshold can be adjusted according to the performance requirements of the model and the application scenario. For example, it can be set to 0.05 or 0.1.

[0060] Preferably, when implementing the feedback optimization module, the operation steps of each unit can be further refined or alternative solutions can be provided. For example, the parameter correction unit can adopt an adaptive adjustment strategy to dynamically adjust the penalty term weight according to the frequency and error magnitude of the decision execution result, rather than simply making a reverse adjustment. The weight decay unit can perform differential decay in combination with the importance of features, using a lower decay rate for important features and a higher decay rate for secondary features. The version iteration unit can introduce a model performance evaluation mechanism to comprehensively evaluate the current model before generating a new model version to ensure that the generation of the new version is necessary and effective. In addition, a model version rollback mechanism can be set up so that when the performance of the new version is inferior to that of the old version, it can be rolled back to the old version in a timely manner to ensure the stability and reliability of the system.

[0061] In some embodiments, the spatial distribution analysis includes: Identifying data abnormal aggregation regions through a density clustering algorithm with a dynamically adjusted neighborhood radius; Generating a three-dimensional visualization map of the feature distribution density; Establishing a pathological feature transfer matrix between different spatial regions.

[0062] Specifically, the data abnormal aggregation regions can be identified through an improved DBSCAN algorithm; Generating a three-dimensional visualization map according to the feature distribution density; Establishing a pathological feature transfer matrix between different spatial regions.

[0063] It should be noted that the spatial distribution analysis in the multi-dimensional analysis module is an important part of this system, aiming to reveal the spatial distribution characteristics and potential laws of data through various analysis means. This analysis includes a density clustering unit, a heat map generation unit, and an association mapping unit. The density clustering unit identifies abnormally aggregated data regions through an improved DBSCAN algorithm. The heat map generation unit generates a three-dimensional visualization map according to the feature distribution density. The association mapping unit establishes a pathological feature transfer matrix between different spatial regions to provide intuitive data support for clinical decision-making.

[0064] Specifically, the DBSCAN algorithm adopted by the density clustering unit is a density-based clustering algorithm that can effectively identify clusters of any shape. The improved DBSCAN algorithm better adapts to data distributions of different densities by dynamically adjusting the neighborhood radius. The heat map generation unit generates a three-dimensional visualization map according to the feature distribution density, where the color or height represents the data density, helping users visually observe the data distribution. The transfer matrix established by the association mapping unit is used to describe the pathological feature changes between different spatial regions, and the elements in the matrix represent the feature transfer probability or intensity from one region to another.

[0065] Preferably, when implementing the spatial distribution analysis, the operation steps of each unit can be further refined or alternative solutions can be provided. For example, the improved DBSCAN algorithm of the density clustering unit can be combined with other clustering algorithms, such as K-means or Gaussian Mixture Models, to improve the accuracy and robustness of clustering. The heat map generation unit can adopt different visualization techniques, such as two-dimensional heat maps, contour maps, or three-dimensional surface maps, to adapt to different data characteristics and user needs. The transfer matrix of the association mapping unit can be combined with time series information to construct a dynamic transfer matrix to capture the changing trends of pathological features over time. In addition, machine learning algorithms, such as random forests or support vector machines, can be introduced to further analyze and predict the clustering results and transfer matrices to provide deeper insights and decision-making support.

[0066] In some embodiments, the density clustering algorithm includes: Dynamically adjust the neighborhood radius according to the average distance, distance standard deviation, and number of data points: Adopt a two-way linking strategy to merge overlapping clusters; Determine the probability attribution of border points.

[0067] Furthermore, the following formula is used to dynamically adjust the neighborhood radius:

[0068] Wherein, is the average distance, is the distance standard deviation, and N is the number of data points.

[0069] It should be noted that the improved DBSCAN algorithm is the core technology of spatial distribution analysis in the multi-dimensional analysis module, which is used to identify abnormal clustering regions in data. By dynamically adjusting the neighborhood radius, adopting a two-way link strategy to merge overlapping clusters, and making a probabilistic membership determination for border points, the accuracy and robustness of clustering are improved. In the formula for dynamically adjusting the neighborhood radius, the average distance and the distance standard deviation are used to measure the distribution density of data points, and the number of data points is used to adjust the size of the neighborhood radius to adapt to clustering regions with different densities. The two-way link strategy effectively merges overlapping clusters by establishing two-way links during the clustering process, avoiding duplicate allocation of data points. The probabilistic membership determination assigns a probability value to border points, indicating the probability that they belong to a certain cluster, thereby improving the reliability of the clustering results.

[0070] Specifically, in the formula for dynamically adjusting the neighborhood radius, the average distance can be obtained by calculating the average Euclidean distance between all data points, and the distance standard deviation is obtained by calculating the standard deviation of the distances between all data points. The number of data points refers to the number of data points in the current clustering region. In the two-way link strategy, each data point establishes links with other data points during the clustering process. When there is an overlap between two clustering regions, they are merged into one cluster through two-way links. In the probabilistic membership determination, the probability value of a border point can be determined by calculating its distance from the cluster center and the distance distribution of other data points within the cluster. The closer the probability value is to 1, the more likely the border point belongs to the cluster.

[0071] Preferably, when implementing the improved DBSCAN algorithm, each step can be further refined or alternative solutions can be provided. For example, when dynamically adjusting the neighborhood radius, the local density characteristics of the data can be combined, and a local density estimation method can be used to calculate the neighborhood radius to better adapt to clustering regions with different densities. In the two-way link strategy, a weight mechanism can be introduced to assign weights to links according to the similarity between data points, so as to better consider the similarity of data points when merging overlapping clusters. In the probabilistic membership determination, a Bayesian method can be used to calculate the posterior probability that a data point belongs to a certain cluster based on its prior probability and likelihood probability to improve the accuracy of the determination. In addition, other clustering algorithms, such as K-means or Gaussian Mixture Models, can be combined to post-process the results of the DBSCAN algorithm to further improve the accuracy and robustness of clustering.

[0072] In some embodiments, the real-time storage unit adopts the following index structure: Time partition layer: Divide the data storage block by hour granularity; Feature Encoding Layer: Append a feature hash value to each piece of data; Fast Retrieval Layer: Build a cross-partition joint query index based on the B+ tree.

[0073] It should be noted that the real-time storage unit is responsible for efficiently storing and managing multi-source medical data in this system, and the index structure it adopts is crucial for the fast retrieval and analysis of data. This index structure includes a time partition layer, a feature encoding layer, and a fast retrieval layer. The time partition layer divides the data storage blocks by hour granularity to ensure the ordered storage of time series data. The feature encoding layer appends a feature hash value to each piece of data to quickly identify and locate data with the same features. The fast retrieval layer builds a cross-partition joint query index based on the B+ tree to achieve efficient data retrieval and query operations.

[0074] Specifically, the hour granularity division of the time partition layer means that the data will be stored and indexed according to each hourly time period. For example, all data from 2024-06-01 00:00:00 to 2024-06-01 00:59:59 will be stored in the same block. This division method helps to improve the retrieval efficiency of time series data, especially when performing time series analysis. The feature encoding layer encodes the feature information of the data into a fixed-length string by generating a feature hash value for each piece of data. This not only helps to quickly identify the data but also reduces the storage space occupancy when the data volume is large. The fast retrieval layer uses the index structure of the B+ tree to achieve fast query of data across multiple time partitions. Each node of the B+ tree contains multiple key values and pointers to child nodes, and this structure is particularly suitable for the storage and retrieval of large amounts of data.

[0075] Preferably, when implementing the real-time storage unit, the parameters and operation steps of each layer can be further refined or alternative solutions can be provided. For example, the granularity of the time partition layer can be adjusted according to the actual data volume and query requirements, such as dividing by day or by minute. Different algorithms can be selected for the hash function of the feature encoding layer, such as MD5, SHA-1, or SHA-256, to ensure the uniqueness and security of the hash value. The B+ tree structure of the fast retrieval layer can be optimized in combination with the distribution characteristics of the data, such as adjusting the order of the tree or the size of the nodes, to improve the retrieval performance. In addition, a caching mechanism can be introduced to cache frequently accessed data to further improve the data retrieval speed.

[0076] The above-mentioned embodiments of the present invention have the following beneficial effects: The system can effectively integrate multi-source heterogeneous medical data, accurately extract multi-dimensional features and train an adaptive decision-making model through dynamic weight allocation and optimization algorithms, improving the accuracy and reliability of medical data analysis. The system can perform spatial distribution, temporal evolution and correlation analysis, providing comprehensive and scientific basis for clinical decision-making, and effectively assisting medical staff in formulating more accurate treatment plans.

[0077] The system can also verify the reliability of the decision through multi-stage logical judgment based on the actual application feedback, and dynamically adjust the model parameters and weight allocation strategy to continuously optimize the system performance. At the same time, the system has efficient data collection, preprocessing and storage capabilities, can filter invalid data in real time, fill in missing values, perform standardization processing, and establish a distributed index storage structure to ensure the integrity and availability of data, providing strong support for medical data analysis and decision-making.

[0078] Embodiment 2: Please refer to Figure 2 , Figure 2 which is a schematic flowchart of a multi-dimensional medical data analysis and auxiliary decision-making method provided by an embodiment of the present invention. As Figure 2 shown, a specific embodiment of the present invention discloses a multi-dimensional medical data analysis and auxiliary decision-making method, including the following steps: S100, obtaining structured and unstructured data from multi-source medical devices and databases; S200, performing noise filtering, missing value filling and standardization processing on the obtained structured and unstructured data; S300, extracting multi-dimensional features from the preprocessed data through a dynamic weight allocation algorithm; S400, training an adaptive decision-making model based on the extracted multi-dimensional features and updating the model parameters using a dynamic weight optimization algorithm; S500, performing spatial distribution analysis, temporal evolution analysis and correlation analysis on the feature data; S600, generating clinical decision-making suggestions based on the multi-dimensional analysis results and verifying the decision reliability through multi-stage logical judgment; S700, adjusting the model parameters and weight allocation strategy according to the actual application feedback.

[0079] It can be understood that the steps described in the multi-dimensional medical data analysis and auxiliary decision-making method correspond to the respective modules in the multi-dimensional medical data analysis and auxiliary decision-making system described in reference Figure 1 . Therefore, the modules, features and beneficial effects described above for the multi-dimensional medical data analysis and auxiliary decision-making system also apply to the multi-dimensional medical data analysis and auxiliary decision-making method and the operations included therein, and will not be elaborated herein.

[0080] Furthermore, the storage medium of the embodiment of the present application stores program instructions that can implement all of the above methods. Among them, the program instructions can be stored in the above storage medium in the form of a software product, including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods described in various embodiments of the present application. The foregoing storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs, or terminal devices such as computers, servers, mobile phones, and tablets.

[0081] The above description is only some preferred embodiments of the present invention and an explanation of the technical principles applied. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present invention is not limited to the technical solutions formed by the specific combination of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the technical solutions formed by mutually replacing the above features with technical features having similar functions (but not limited to) disclosed in the embodiments of the present invention.

Claims

1. A multi-dimensional medical data analysis and auxiliary decision-making system, characterized in that It includes the following modules: A data acquisition module, which is used to obtain structured and unstructured data from multi-source medical devices and databases; A data preprocessing module, which is used to perform noise filtering, missing value filling, and normalization processing on the data output by the data acquisition module; A feature extraction module, which is used to extract multi-dimensional features from the preprocessed data through a dynamic weight allocation algorithm; A model training module, which is used to train an adaptive decision-making model based on the extracted multi-dimensional features and update the model parameters using a dynamic weight optimization algorithm; A multi-dimensional analysis module, which is used to perform spatial distribution analysis, temporal evolution analysis, and correlation analysis on the feature data; A decision generation module, which is used to generate clinical decision-making suggestions based on the multi-dimensional analysis results and verify the decision reliability through multi-stage logical judgment; A feedback optimization module, which is used to adjust the model parameters and weight allocation strategy according to the actual application feedback; The feedback optimization module includes: A parameter correction unit, which reversely adjusts the penalty term weight of the model loss function according to the decision execution result; A weight decay unit, which exponentially decays the feature weights that have not been updated for a long time according to the decay rate and the non-update time interval; A version iteration unit, which generates a new model version when the cumulative error exceeds the preset threshold.

2. The system according to claim 1, wherein The data acquisition module includes: A multi-source interface unit, which is used to connect to the electronic medical record system, medical imaging equipment, and wearable sensors; A dynamic cleaning unit, which is used to filter invalid data in real time based on the preset data quality threshold; A real-time storage unit, which is used to establish a distributed index storage structure according to the data type and timestamp.

3. The system according to claim 1, characterized in that, The data preprocessing module includes: A temporal alignment unit, which is used to interpolate and align multi-source asynchronous data according to a unified time reference; An anomaly correction unit, which is used to dynamically correct the data outside the range of the mean plus or minus three standard deviations of the current data window. The corrected data is generated according to the relationship between the dynamic adjustment coefficient and the mean and standard deviation; A normalization unit, which is used to map the data to the interval from zero to one.

4. The system according to claim 3, characterized in that, The specific steps of the dynamic weight allocation algorithm include: Calculating the sum of the absolute values of the correlation coefficients between each feature dimension and the clinical indicators, and dividing it by the maximum absolute value of the correlation coefficient to obtain the feature importance score; According to the feature importance score, dynamically allocate weights through exponential function normalization; Inputting the weighted features into a convolutional attention network for fusion.

5. The system according to claim 1, characterized in that, The parameter update process of the dynamic weight optimization algorithm includes: Based on the adaptive learning rate, momentum factor, and regularization coefficient, dynamically adjust the model parameters in combination with the historical parameter changes and the gradient of the loss function.

6. The system according to claim 1, wherein The multi-stage logical judgment includes: An initial judgment stage, which matches the preset clinical path rules; A confidence evaluation stage, which calculates the decision confidence through the average difference between the model prediction probability and the historical verification labels; An expert rule verification stage, which calls the expert knowledge base for secondary verification when the confidence is lower than the preset threshold.

7. The system according to claim 1, characterized in that, The spatial distribution analysis includes: Identifying data abnormal aggregation areas through a density clustering algorithm with a dynamically adjusted neighborhood radius; Generating a three-dimensional visualization map of the feature distribution density; Establishing a pathological feature transfer matrix between different spatial regions.

8. The system according to claim 7, wherein The density clustering algorithm includes: Dynamically adjust the neighborhood radius according to the average distance, distance standard deviation, and number of data points; Adopt a two-way link strategy to merge overlapping clusters; Determine the probability attribution of boundary points.

9. A multi-dimensional medical data analysis and auxiliary decision-making method, characterized in that, It includes the following steps: Obtain structured and unstructured data from multi-source medical devices and databases; Perform noise filtering, missing value filling, and standardization processing on the obtained structured and unstructured data; Extract multi-dimensional features from the preprocessed data through a dynamic weight allocation algorithm; Train an adaptive decision model based on the extracted multi-dimensional features, and update the model parameters using a dynamic weight optimization algorithm; Conduct spatial distribution analysis, temporal evolution analysis, and correlation analysis on the feature data; Generate clinical decision recommendations based on the multi-dimensional analysis results, and verify the reliability of the decisions through multi-stage logical judgment; Adjust the model parameters and weight allocation strategy according to the actual application feedback.

Images