Tuberculosis latent infection deep learning assisted prevention and treatment management platform

Abstract

The application discloses a tuberculosis latent infection deep learning auxiliary prevention and treatment management platform, relates to the technical field of tuberculosis latent infection auxiliary prevention, and comprises a data acquisition and dynamic monitoring module, a data preprocessing and organization module, a conflict feature extraction and analysis module, a deep learning conflict evaluation module and a dynamic weight adjustment and optimization module.The data acquisition and dynamic monitoring module acquires fusion parameter information of multi-modal data in real time, ensures that each input can be dynamically monitored in the data fusion process, and thus provides basic data for subsequent conflict detection, feature extraction and intelligent evaluation.The application extracts conflict features and intelligently evaluates data source conflicts by means of a deep learning model, dynamically adjusts weights, and optimizes the data fusion process by monitoring multi-modal data fusion parameters in real time.The scheme improves the accuracy of tuberculosis latent infection risk evaluation, avoids misdiagnosis and missed diagnosis, and promotes the implementation of precision medicine and individualized treatment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of latent tuberculosis infection auxiliary prevention, and specifically relates to a latent tuberculosis infection deep learning auxiliary prevention and treatment management platform. BACKGROUND

[0002] Latent Tuberculosis Infection (LTBI) refers to the infection of Mycobacterium tuberculosis after the human immune system can control and inhibit the activity of the bacteria, and the infected person has no obvious clinical symptoms, but may be converted into active tuberculosis when the immunity is decreased. Latent Tuberculosis Infection deep learning auxiliary prevention and treatment management refers to using deep learning technology to analyze and model based on a large amount of patient data (such as imaging data, genomic data, clinical records, etc.) to assist in the early screening, risk assessment, treatment decision and management optimization of latent tuberculosis infection. Through deep learning algorithms, the risk factors of latent infection can be automatically identified, the possibility of conversion of latent infection into active tuberculosis can be predicted, and accurate prevention and treatment strategies can be recommended according to individual health conditions. This method can improve the accuracy and efficiency of latent tuberculosis infection management, reduce the risk of conversion into active tuberculosis, and thus help control the spread of tuberculosis.

[0003] The prior art has the following disadvantages: The prior art usually fuses these data into comprehensive features to comprehensively evaluate the risk of conversion of latent tuberculosis infection into active tuberculosis when analyzing based on a large amount of patient data such as imaging data, genomic data and clinical records. However, when the features of different data sources conflict, the model may have difficulty in accurately processing. For example, imaging data shows that the lung lesion is slight, while genomic data or clinical records show that the patient has a higher susceptibility. This conflict may cause the model to deviate in the weight distribution of these features, thereby affecting the final prediction result. If this conflict is not effectively handled, the model may overestimate or underestimate the conversion risk, further leading to unnecessary long-term treatment, drug side effects, or neglect of low-risk patients, ultimately missing the opportunity for early intervention, leading to missed diagnosis of active tuberculosis, and thus affecting public health safety.

[0004] The above information disclosed in the BACKGROUND section only serves to enhance the understanding of the background of the present disclosure, and thus it can include information that does not constitute the prior art known to those of ordinary skill in the art. SUMMARY

[0005] The purpose of the present application is to provide a latent tuberculosis infection deep learning assisted prevention and treatment management platform, which can monitor multi-modal data fusion parameters in real time, extract conflict features by feature engineering, and intelligently evaluate data source conflicts by a deep learning model, dynamically adjust the weight of each data source, and optimize the data fusion process. This scheme effectively improves the accuracy and robustness of latent tuberculosis infection conversion risk assessment, avoids misdiagnosis and missed diagnosis, and promotes the implementation of precision medicine and personalized treatment, to solve the problems in the above background art.

[0006] To achieve the above purpose, the present application provides the following technical scheme: a latent tuberculosis infection deep learning assisted prevention and treatment management platform, comprising a data acquisition and dynamic monitoring module, a data preprocessing and organization module, a conflict feature extraction and analysis module, a deep learning conflict evaluation module, and a dynamic weight adjustment and optimization module:

[0007] The data acquisition and dynamic monitoring module acquires fusion parameter information of multi-modal data in real time, ensures dynamic monitoring of each input during data fusion, and provides basic data for subsequent conflict detection, feature extraction, and intelligent evaluation;

[0008] The data preprocessing and organization module organizes the acquired multi-modal data fusion parameter information into a unified data set and pre-processes the multi-modal data fusion parameter information in the data set, facilitating subsequent analysis;

[0009] The conflict feature extraction and analysis module extracts key features reflecting conflicts in data fusion from the pre-processed data through feature engineering technology, and analyzes the extracted key features, providing strong support for subsequent conflict detection;

[0010] The deep learning conflict evaluation module inputs the analyzed key features into a pre-learned deep learning model, and identifies whether there is a conflict in the data fusion process through the learning model;

[0011] The dynamic weight adjustment and optimization module dynamically adjusts the weight of each data source according to its reliability and conflict severity after detecting conflicts in the data fusion process and evaluating their severity, reduces the negative impact of conflict data sources on the final decision, and thus improves the accuracy and robustness of the fusion result.

[0012] Preferably, the specific steps of organizing the acquired multi-modal data fusion parameter information into a unified data set are as follows: first, label and normalize the fusion parameters of each data source to conform to a unified data structure;

[0013] Then, denoise, handle missing values, and detect anomalies to eliminate potential quality problems;

[0014] Subsequently, the data is uniformly scaled through standardization techniques to ensure the comparability and consistency of parameters from different data sources in subsequent analysis;

[0015] Finally, the pre-processed data set is used for further feature extraction, conflict detection and input of deep learning models to ensure the accuracy and reliability of the analysis results.

[0016] Preferably, key features reflecting conflicts in data fusion are extracted from the pre-processed data through feature engineering techniques. The extracted features include the contribution change of each data source to the final decision in the multi-modal fusion process and the error change in the multi-modal data fusion process. The contribution change of each data source to the final decision in the multi-modal fusion process and the error change in the multi-modal data fusion process are analyzed under the detection window to generate data source contribution change reference values and data fusion error reference values, respectively. The data source contribution change reference values are used to quantify the influence and contribution change of each data source in the multi-modal data fusion process, and the data fusion error reference values are used to measure the inconsistency in the multi-modal data fusion process over time.

[0017] Preferably, the specific steps for generating data source contribution change reference values by analyzing the contribution change of each data source to the final decision in the multi-modal fusion process under the detection window are as follows:

[0018] In the multi-modal data fusion process, the contribution of each data source to the final decision will be continuously adjusted as the data input changes. To quantify the contribution of the data source at each time point, a dynamic weight function is introduced, which is calculated by considering the feature importance of the data source and the response of the fusion model. The calculation expression is as follows: ,

[0019] In the formula, is the contribution weight of the th data source to the decision at time , is the feature intensity of the th data source at time , is the dynamic adjustment factor related to the th data source at time , is the feature intensity of the th data source at time , is the dynamic adjustment factor related to the th data source at time , is the total number of all data sources used in the multi-modal data fusion process;

[0020] Next, by performing time series analysis on the dynamic contribution of each data source, the change in the contribution of each data source is calculated, a contribution change index is introduced to quantify the change in the contribution of the data source between the current time and the previous time, and the calculation expression is as follows: ,

[0021] In the formula, is the contribution change index of the i-th data source at time t, represents the contribution change index of the i-th data source at time t, is the contribution weight of the i-th data source to the decision at time t, is the conflict adjustment factor of the i-th data source, is a time-dependent factor representing the influence of external environmental changes on the i-th data source at the current time;

[0022] Finally, the conflict of all data sources in the multi-modal data fusion is evaluated as a whole, the contribution change indexes of the data sources are fused to generate a data source contribution change reference value, and the calculation expression is as follows: ,

[0023] In the formula, is the data source contribution change reference value.

[0024] Preferably, the error change generated in the multi-modal data fusion process is analyzed under the detection window to generate the specific steps of the data fusion error reference value as follows:

[0025] In the first step of generating the data fusion error index, the error of each data source is first calculated, and the error of each data source represents the deviation of the data source to the final prediction result, and the error calculation expression is as follows: ,

[0026] In the formula, is the error of the i-th data source, is the prediction result of the i-th data source on the j-th sample, is the actual value of the j-th sample, is the error amplification index of the i-th data source, is the total number of samples in the data set, i.e., the number of samples processed by the model when making predictions;

[0027] ​​​​​​​​​​​​​​​After calculating the errors from each data source, the obtained errors are then weighted and fused to obtain a comprehensive error value. The error from each data source is weighted according to its importance to the fusion result, and the calculation expression is as follows: ,

[0028] In the formula, It is the total error after merging all data sources. It is the first The weight of each data source, It is the first Error amplification index of each data source, It represents the total number of all data sources used in the multimodal data fusion process;

[0029] Finally, the fused error value is transformed into a data fusion error reference value to quantify the degree of conflict in the multimodal data fusion process. A nonlinear function is used to map the fused error, and the calculation expression is as follows: ,

[0030] In the formula, This is a reference value for data fusion error. It is the first Weight coefficients of each data source, It is the first Weighted nonlinear transformation of errors from multiple data sources.

[0031] Preferably, the analyzed reference values ​​for changes in data source contributions and reference values ​​for data fusion errors are input into a pre-learned deep learning model. The deep learning model generates a fusion conflict risk coefficient, which is then used to intelligently evaluate the multimodal data fusion process and identify whether conflicts exist during the data fusion process.

[0032] Preferably, a pre-learned deep learning model is used to intelligently evaluate the multimodal data fusion process. When conflicts occur during the data fusion process, the generated fusion conflict risk coefficient is compared and analyzed with a pre-set fusion conflict risk coefficient to divide the multimodal data fusion process. The division steps are as follows:

[0033] If the fusion conflict risk coefficient is greater than the preset reference threshold for fusion conflict risk coefficient, the current multimodal data fusion process is classified as having data fusion conflict; if the fusion conflict risk coefficient is less than or equal to the preset reference threshold for fusion conflict risk coefficient, the current multimodal data fusion process is classified as not having data fusion conflict.

[0034] Preferably, after detecting conflicts during the data fusion process and assessing their severity, the specific steps for dynamically adjusting the weights based on the reliability and conflict severity of each data source are as follows:

[0035] After detecting conflicts during the data fusion process, the severity of the conflicts needs to be assessed first, and the reliability of each data source needs to be quantitatively evaluated. The evaluation formula is as follows: ,

[0036] In the formula, It is a data source The conflict severity index represents the data source. The severity of conflicts arising during multimodal data fusion. It is a data source The conflict sensitivity parameter is adjusted to determine its responsiveness to conflict. It is a coefficient of risk of fusion conflict. This is a reference threshold for the risk coefficient of fusion conflict. It is a data source The exponential growth factor;

[0037] When assessing the reliability of the data source, a multi-weighted standardization approach is used, incorporating historical performance weight adjustments. This approach considers the weighted impact of data quality and historical performance, and balances importance using a composite function. The calculation expression is as follows: ,

[0038] In the formula, It is a data source Reliability rating It is a data source Quality rating It is a data source Historical performance rating This is the stability coefficient for each data source, reflecting the weighted impact of historical performance on reliability assessment. It is a data source Historical variability indicates the stability of the data source throughout history. It is the mean of the historical variability across all data sources. and These are the data sources. Quality rating and data source Historical performance score Weighting coefficients;

[0039] In the final dynamic weight adjustment, in addition to adjusting for conflict severity and reliability, a composite weighted adjustment function is introduced. This function incorporates exponential weighting, considers the interaction between conflict severity and data reliability, and also includes enhanced contrast adjustment.

[0040] Based on the assessment results of conflict severity and data source reliability, the weight of each data source is dynamically adjusted, and the calculation formula is as follows: ,

[0041] In the formula, It is the adjusted number The weight of each data source, These are the original weights. It is an adjustment coefficient that controls the data source. Adjusting sensitivity, These are non-linear weighting coefficients for reliability scoring. This is a conflict severity adjustment factor that controls the impact of conflict on weight adjustments. It is a weighted index of conflict severity, which modulates the nonlinear effect of conflict severity on the weight.

[0042] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0043] This invention effectively solves the data conflict problem in the multimodal data fusion process by acquiring and dynamically monitoring multimodal data fusion parameters in real time, thereby improving the accuracy and robustness of risk assessment for the conversion of latent tuberculosis infection into active tuberculosis. Specifically, key features reflecting data conflicts are extracted through feature engineering techniques, and intelligent evaluation is performed using a pre-learned deep learning model to promptly identify conflicts between data sources. After a conflict is detected, the weights of each data source are dynamically adjusted based on the reliability and severity of the conflict, reducing the negative impact of conflicting data sources on the final decision. This approach not only optimizes the accuracy of the multimodal data fusion process and avoids the risks of misdiagnosis and missed diagnosis, but also enhances the model's adaptability in complex clinical data environments, ultimately promoting the implementation of precision medicine and personalized treatment. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0045] Figure 1 This is a schematic diagram of the modules of the deep learning-assisted prevention and treatment management platform for latent tuberculosis infection of the present invention. Detailed Implementation

[0046] Example implementations will now be described more fully with reference to the accompanying drawings. Example implementations, however, can be implemented in many different forms and should not be construed as limited to the examples set forth herein; rather, these example implementations are provided so that this disclosure will be thorough and complete, and will fully convey the inventive aspects to those skilled in the art.

[0047] The present application provides a deep learning assisted prevention and treatment management platform for latent tuberculosis infection as shown in Figure 1 The deep learning assisted prevention and treatment management platform for latent tuberculosis infection includes a data acquisition and dynamic monitoring module, a data preprocessing and organization module, a conflict feature extraction and analysis module, a deep learning conflict evaluation module, and a dynamic weight adjustment and optimization module.

[0048] The data acquisition and dynamic monitoring module acquires fusion parameter information of multi-modal data in real time, ensures dynamic monitoring of each input during data fusion, and thus provides basic data for subsequent conflict detection, feature extraction, and intelligent evaluation.

[0049] In multi-modal data fusion, the quality and features of each data source such as imaging data, genomic data, and clinical records may change over time or at different stages of data collection. Real-time acquisition of fusion parameter information includes capturing multi-dimensional information such as quality score, integrity, timeliness, and impact on the model of each data source.

[0050] The data preprocessing and organization module organizes the acquired multi-modal data fusion parameter information into a unified data set and pre-processes the multi-modal data fusion parameter information in the data set for subsequent analysis.

[0051] When organizing the acquired multi-modal data fusion parameter information into a unified data set, the parameters from different data sources first need to be formatted and standardized to ensure that they can be compared and analyzed under the same framework. The specific steps include: first, labeling and normalizing each data source's fusion parameters (such as the resolution of imaging data, the accuracy of genomic data, and the timeliness of clinical data) to conform to a unified data structure, such as a table, time series, or data stream. Next, perform noise reduction, missing value processing, and anomaly detection to eliminate potential quality issues such as missing fields, incorrect measurements, or inconsistent timestamps. Then, perform uniform scale adjustment on the data through standardization techniques such as Z-score standardization or Min-Max scaling to ensure that the parameters of different data sources are comparable and consistent in subsequent analysis. The pre-processed data set is used for further feature extraction, conflict detection, and input of the deep learning model, ensuring the accuracy and reliability of the analysis results.

[0052] The establishment of data set is to integrate parameter information from different data sources (such as image, genome, clinical record, etc.) into a standardized and structured form for subsequent processing. These information may include the characteristic types of each data source (such as the resolution of image data, the accuracy of genome data, the timeliness of clinical data, etc.), as well as the relationship or mutual influence between them. The key of this step is to make the parameters accessible conveniently in the subsequent steps through data arrangement and unified formatting, and to provide sufficient data support for conflict detection and feature extraction.

[0053] Data preprocessing is a very critical step in machine learning and deep learning, which involves removing redundant information, handling missing values, standardizing data, removing noise, etc. For example, clinical data may have missing values, image data may have noise, and genomic data may have sequencing errors, etc. The purpose of preprocessing is to unify the format of data and make it have better learning characteristics. In this stage, techniques such as missing value imputation, normalization, data smoothing and denoising may be applied to ensure the quality of data entering the feature extraction and analysis stage does not affect the final fusion result.

[0054] The conflict feature extraction and analysis module extracts key features reflecting the conflict of data fusion from the preprocessed data through feature engineering technology, and analyzes the extracted key features, which provides strong support for subsequent conflict detection;

[0055] The conflict feature extraction and analysis module extracts key features reflecting the conflict of data fusion from the preprocessed data through feature engineering technology, and analyzes the extracted key features, which provides strong support for subsequent conflict detection;

[0056] When the contribution of each data source to the final decision changes exponentially during multimodal data fusion, it usually indicates potential data conflict in the current multimodal data fusion process. This is because, in a normal fusion process, the contributions between data sources should be relatively stable unless the performance of a particular data source undergoes a sudden change or its association with other data sources is significantly altered. If the contribution of a particular data source increases exponentially, it may mean that there is significant inconsistency or anomaly between that data source and other data sources. For example, imaging data may suddenly become overly important at a certain moment due to technical glitches or quality issues, causing the model to over-rely on that data source in its decision-making; or, the contribution of genomic data may increase sharply at a certain moment, indicating that its expected relationship with imaging or clinical data has been broken, thus affecting the fusion results. Conversely, a sharp decline in the contribution of some data sources may also indicate that the synergy between data sources has been disrupted, or that the importance of that data source in the fusion has been underestimated. Therefore, exponential changes usually reveal potential conflicts between data sources, which may lead the model to make incorrect predictions or decisions, thereby affecting the final accuracy and stability.

[0057] The specific steps for analyzing the contribution changes of each data source to the final decision during the multimodal fusion process and generating reference values ​​for the contribution changes of data sources within the detection window are as follows:

[0058] In the process of multimodal data fusion, the contribution of each data source to the final decision is constantly adjusted as the data input changes. To quantify the contribution of each data source at each time point, a dynamic weighting function is introduced. This function is calculated by considering the feature importance of the data source and the response of the fusion model. The calculation expression is as follows: ,

[0059] In the formula, It is the first Data sources in time The weight of each time's contribution to the decision-making process. It is the first Data sources at time The strength of features (e.g., data quality, signal-to-noise ratio, etc.) It is time The first time The dynamic adjustment factors associated with each data source reflect the influence of that data source on the final decision. It is the first Data sources at time Feature intensity on, It is time The first time Dynamic adjustment factors related to each data source It represents the total number of all data sources used in the multimodal data fusion process;

[0060] The purpose of this step is to provide basic data for the subsequent calculation of contribution changes.

[0061] Next, by performing time series analysis on the dynamic contribution of each data source, the change in the contribution of each data source is calculated, and a contribution change index is introduced to quantify the contribution change of the data source between the current time and the previous time. The calculation expression is as follows:

[0062] In the formula, is the contribution change index, which represents the contribution change of the th data source at time , is the contribution weight of the th data source to the decision at time , is the conflict adjustment factor of the th data source, is a time-dependent factor representing the influence of external environmental changes on the th data source at the current time;

[0063] The role of the above step is to increase the contribution of those that change dramatically, highlighting the abnormal situation that may exist, thereby helping to discover potential conflicts.

[0064] Finally, the conflict situation of all data sources in the multi-modal data fusion is evaluated as a whole, and the contribution change indexes of each data source are fused to generate a data source contribution change reference value, and the calculation expression is as follows:

[0065] In the formula, is the data source contribution change reference value.

[0066] By weighted summing the contribution change indexes of each data source, a comprehensive conflict reference value can be obtained. If the reference value is large, it indicates that there is a significant conflict in the current multi-modal data fusion process; otherwise, it indicates that the contributions among the data sources are consistent, and the fusion process is normal.

[0067] ​​The greater the data source contribution change reference value generated after analyzing the contribution change of each data source to the final decision in the detection window in the multi-modal fusion process, generally means that there is a data conflict in the current multi-modal data fusion process. When the contribution of a certain data source changes significantly, especially in an exponential manner, it often indicates that the relationship between the features of that data source and other data sources is not coordinated or abnormal. A larger data source contribution change reference value reflects a dramatic change in the influence of that data source on the final decision, which may be caused by data quality problems, inherent inconsistencies between data sources, or incorrect weight allocation. Such dramatic changes often indicate conflicts in certain data sources in multi-modal fusion, which can cause instability and increased errors in model decision-making. Conversely, when the data source contribution change reference value is small and stable, it indicates that the interaction between each data source is consistent, and the model can better fuse different data sources, and the decision-making process is stable without significant data conflicts.

[0068] Exponential increase in errors generated in the multi-modal data fusion process is often an important indication of data conflict. When fusing data from different sources, the model integrates information from each data source to make predictions or decisions. If the error of a certain data source increases exponentially during the fusion process, it may indicate that there is a significant inconsistency or bias between that data source and other data sources. As the error gradually increases and the growth rate increases exponentially, it indicates that the contradictions between different data sources have not been effectively handled, which in turn leads to a serious decline in the stability and accuracy of the fusion results. At this time, the conflict of certain features or parameters becomes more prominent, which may be due to the incorrect modeling of the relationship between data sources or unreasonable weight allocation. For example, imaging data and genomic data show inconsistent features in the same patient (such as no obvious lesions in imaging, but genomic data shows that the patient has a higher susceptibility), which will make the model unable to find a reasonable balance point between the two types of information, resulting in rapid accumulation of errors. This exponentially increasing error indicates that a conflict has occurred in the data fusion process, affecting the accuracy of the overall decision, and may even lead to inaccurate risk assessment and diagnostic results. Therefore, when the error index increases exponentially, it not only reflects the conflict between data sources, but also may indicate that the model has failed to correctly handle this conflict, thereby severely affecting the fusion results.

[0069] The specific steps for generating the data fusion error reference value by analyzing the error change in the multi-modal data fusion process under the detection window are as follows:

[0070] In the first step of generating the data fusion error index, the error of each data source is first calculated, and the calculated error will serve as the basis for subsequent calculations. The error of each data source represents the deviation of that data source from the final prediction result, and the error calculation expression is as follows: ,

[0071] In the formula, It is the first Error from each data source It is in the The first sample Prediction results from each data source It is the first The actual value of a sample represents the true situation of that sample. It is the first Error amplification index of each data source, It is the total number of samples in the dataset, that is, the number of samples processed by the model when making predictions;

[0072] The purpose of this step is to provide foundational data for subsequent collision detection by calculating the error of each data source and amplifying it non-linearly.

[0073] After calculating the errors from each data source, the acquired errors are then weighted and fused to obtain a comprehensive error value. This error value directly affects the generation of the final error reference value. The errors from each data source are weighted according to their importance to the fusion result. The weights can be obtained through a pre-defined reliability assessment or through dynamic learning. The calculation expression is as follows: ,

[0074] In the formula, It is the total error after merging all data sources. It is the first The weight of each data source represents the proportion that data source accounts for in the final decision. It is the first The error amplification index of a data source indicates the degree of non-linear adjustment applied to the error of that data source. It represents the total number of all data sources used in the multimodal data fusion process;

[0075] This weighted fusion formula allows us to precisely control the contribution of each data source to the total error based on its reliability and importance. This step ensures the weighting and adjustment of errors during the fusion process, enabling the system to flexibly respond to differences and conflicts between data sources.

[0076] Finally, the fused error value is transformed into a data fusion error reference value to quantify the degree of conflict in the multimodal data fusion process. A nonlinear function is used to map the fused error, and the calculation expression is as follows: ,

[0077] In the formula, This is a reference value for data fusion error. It is the first a weight coefficient of the data source, is the weighted nonlinear transformation of the error of the first

[0078] The formula can effectively identify and quantify the conflicts of different data sources by performing exponential transformation and weighting on the error. Especially when the error of a certain data source increases, the growth of the exponential function will accelerate, thus reflecting a higher conflict level in the error index. Through this method, the system not only identifies the existence of conflicts, but also quantifies the severity of conflicts, helping to adjust and optimize the fusion strategy of the model.

[0079] The larger the data fusion error reference value generated by analyzing the error changes in the detection window during the multi-modal data fusion process, the more likely there is a data conflict in the current multi-modal data fusion process. When multiple data sources (such as imaging data, genomic data, and clinical records) are fused, the model will make decisions based on the weight and features of each data source. If the error value of a certain data source gradually increases during the fusion process, and its growth trend shows an exponential trend, it means that the conflict or inconsistency between different data sources is gradually amplified, which is usually because the features of the data sources cannot work effectively together, leading to the accumulation of error in the model's prediction results. For example, when imaging data and genomic data produce significantly different results in risk assessment for the same patient, the model cannot correctly reconcile the differences between these two data sources, leading to the continuous amplification of errors. On the contrary, if the data fusion error reference value is small and stable, it means that the synergy between different data sources is good, and the model can consistently integrate the information of each data source, indicating that there is no obvious data conflict in the current fusion process.

[0080] The deep learning conflict evaluation module inputs the analyzed key features into a pre-learned deep learning model to identify whether there is a conflict in the data fusion process through the learning model;

[0081] The data source contribution change reference value and the data fusion error reference value after analysis are input into a pre-learned deep learning model to generate a fusion conflict risk coefficient through the deep learning model. The multi-modal data fusion process is intelligently evaluated through the fusion conflict risk coefficient to identify whether there is a conflict in the data fusion process.

[0082] The pre-learned deep learning model refers to a model trained based on specific task requirements using deep learning algorithms (such as neural networks, convolutional neural networks, or recurrent neural networks, etc.) through a large amount of labeled data and multiple rounds of training. These models learn the complex nonlinear relationships between data during the training phase by inputting a large amount of historical data, and can extract key features that affect target prediction (such as the risk prediction of latent tuberculosis infection transforming into active tuberculosis). The pre-learned model has learned how to identify the correlation and potential conflict patterns between different data sources (such as imaging data, genomic data, and clinical records, etc.) based on the input multi-modal data through the training process. Through continuous training, the model can adaptively adjust its internal parameters, enabling it to make more accurate decisions when faced with new input data. In the context of multi-modal data fusion, the pre-learned deep learning model already has the ability to effectively evaluate and judge based on features such as data source contribution changes, error reference values, etc. By inputting these features, the model can evaluate the conflict risk that may exist in the current data fusion process based on previously learned rules.

[0083] Specifically, the deep learning model will use key features such as data source contribution changes and error reference values in historical data, combined with the knowledge accumulated during the training process, to determine whether there is a conflict in the current data fusion process. For example, if certain feature combinations (such as the combination of imaging features and genomic features) often lead to high errors or risks in historical data, the model can identify these patterns and find similar situations in new data. Through this intelligent evaluation, the deep learning model can dynamically adjust and predict the risk of the multi-modal data fusion process, improving the accuracy and robustness of data fusion. The advantage of the pre-learned model is that it can not only handle large amounts of data and complex features, but also adapt to the heterogeneity between data sources and possible conflict situations, timely identify and adjust inconsistent data sources, effectively reduce errors, and improve the final prediction ability.

[0084] The deep learning model is not limited here, and a deep learning model that can comprehensively analyze data source contribution change reference values and data fusion error reference values to generate a fusion conflict risk coefficient can be provided as a specific implementation for the technical solution of the present application;

[0085] The fusion conflict risk coefficient is generated according to the following formula: ,

[0086] In the formula, , are data source contribution change reference values and a data fusion error reference value , and , are both greater than 0.

[0087] The preset proportion coefficients and are parameters for measuring the degree of contribution of different data sources to the data fusion error. These two coefficients correspond to the data source contribution change reference value and the data fusion error reference value respectively, which affect the calculation of the final fusion conflict risk coefficient . Specifically, controls the error influence, controls the error influence. By setting and , the sensitivity of the model to different data sources and fusion errors can be adjusted in practical applications to balance their influence on the final conflict assessment. In actual use, and values are usually determined by experiments or tuning methods to ensure that the model can accurately reflect the true contribution of data sources and fusion errors.

[0088] From the fusion conflict risk coefficient, the greater the data source contribution change reference value generated after analyzing the contribution change of each data source to the final decision in the multimodal fusion process in the detection window, the greater the data fusion error reference value generated after analyzing the error change in the multimodal data fusion process in the detection window, indicating that the greater the fusion conflict risk coefficient generated when the pre-learned deep learning model is used to intelligently evaluate the multimodal data fusion process and identify whether there is a conflict in the data fusion process, the greater the probability of data conflict in the multimodal data fusion process, and vice versa. The probability of data conflict in the multimodal data fusion process is smaller.

[0089] The fusion conflict risk coefficient generated when the pre-learned deep learning model is used to intelligently evaluate the multimodal data fusion process and identify whether there is a conflict in the data fusion process is compared and analyzed with the pre-set fusion conflict risk coefficient, and the multimodal data fusion process is divided, and the division steps are as follows:

[0090] If the fusion conflict risk coefficient is greater than the pre-set fusion conflict risk coefficient reference threshold, the current multimodal data fusion process is divided into data fusion conflict; if the fusion conflict risk coefficient is less than or equal to the pre-set fusion conflict risk coefficient reference threshold, the current multimodal data fusion process is divided into data fusion conflict.

[0091] The dynamic weight adjustment and optimization module adjusts the weights of each data source dynamically based on their reliability (such as data quality, historical performance, etc.) and the severity of the conflict, reducing the negative impact of conflicting data sources on the final decision, thereby improving the accuracy and robustness of the fusion results.

[0092] The dynamic weight adjustment and optimization module plays a crucial role in multi-modal data fusion, especially when there are conflicts. It effectively reduces the negative impact of conflicting data sources on the final decision, improving the accuracy and robustness of the overall decision. Specifically, after detecting conflicts in the multi-modal data fusion process and assessing their severity, the module dynamically adjusts the weights of each data source based on their reliability (such as data quality, historical performance, etc.) and the severity of the conflict. The core of this process lies in "dynamic" and "intelligent", which can flexibly adjust the importance of each data source in the fusion process according to its performance in specific situations, rather than assigning the same weight to each data source.

[0093] For example, when the quality of imaging data is poor or there is significant conflict between it and other data sources (such as genomic data), the dynamic weight adjustment module will automatically reduce the weight of imaging data, reducing its impact on the final decision. Similarly, if the reliability of genomic data is high and it has good historical performance in multiple scenarios, the module will increase the weight of genomic data, allowing it to play a greater role in the decision-making process. In this way, the system not only reduces errors caused by data conflicts, but also ensures that the most reliable data source is used in different situations, ultimately improving the accuracy of the fusion results.

[0094] In addition, the dynamic weight adjustment optimization module also enhances the robustness of the system. When data quality fluctuates or encounters new challenges, it can maintain the stability of the fusion process through adaptive adjustment strategies, avoiding the impact of individual data source defects or biases on the reliability of the entire decision-making system. This self-optimizing and adaptive design enables the multi-modal data fusion system to maintain efficient and accurate decision-making capabilities in complex and changing real-world applications.

[0095] After detecting conflicts in the data fusion process and assessing their severity, the specific steps for dynamic weight adjustment based on the reliability of each data source and the severity of the conflict are as follows:

[0096] After detecting conflicts in the data fusion process, the first step is to assess the severity of the conflict and quantitatively evaluate the reliability of each data source, using the following formula: ,

[0097] wherein, is the conflict severity index of data source is the conflict severity index of data source is the conflict sensitivity parameter of data source is the fusion conflict risk coefficient is the fusion conflict risk coefficient reference threshold is the conflict severity index of data source

[0098] The above step adopts exponential amplification, so that the conflict severity shows nonlinear growth when approaching the reference threshold, reflecting the dynamic amplification effect of conflict, especially for conflict-sensitive data sources. In high-risk situations, the conflict severity index becomes more sensitive and dramatic.

[0099] In evaluating the reliability of data sources, a multiple weighted standardization method is combined, and the weight adjustment of historical performance is added, considering the weighted influence of data quality and historical performance, and balancing the importance through a composite function, the calculation expression is as follows:

[0100] In the formula, is the reliability score of data source is the quality score of data source , measuring the accuracy, completeness, etc. is the historical performance score of data source , considering the performance of the data source in the past prediction is the stability coefficient of each data source, reflecting the weighted influence of historical performance on reliability evaluation is the historical variability of data source , indicating the stability of the data source in history is the mean of the historical variability of all data sources and are the weight coefficients of the quality score of data source and the historical performance score of data source , determining the relative importance of the quality score and the historical performance score in the total reliability score

[0101] ​​​​​This step ensures that the reliability score relies not only on the quality and historical performance of the data source, but also adjusts for data stability through historical variability. Higher historical variability indicates lower stability, but its impact is less significant. This formula guarantees a comprehensive and multi-dimensional consideration of reliability assessment.

[0102] In the final dynamic weight adjustment, in addition to adjusting for conflict severity and reliability, a composite weighted adjustment function is introduced. This function incorporates exponential weighting, considers the interaction between conflict severity and data reliability, and also includes enhanced contrast adjustment.

[0103] Based on the assessment results of conflict severity and data source reliability, the weight of each data source is dynamically adjusted, and the calculation formula is as follows: ,

[0104] In the formula, It is the adjusted number The weight of each data source, These are the original weights, representing the initial contribution of the data source when there are no conflicts. It is an adjustment coefficient that controls the data source. Adjusting sensitivity, It is a non-linear weighting coefficient for the reliability score, representing the degree to which reliability affects the weight adjustment. This is a conflict severity adjustment factor that controls the impact of conflict on weight adjustments. It is a weighted index of conflict severity, which modulates the nonlinear effect of conflict severity on the weight.

[0105] This step comprehensively considers the interactive effects of data source reliability and conflict severity, achieving dynamic optimization of weights through non-linear adjustments to both. When conflicts are severe, the weights decrease exponentially, and conversely, when data source reliability is low, the weights also decrease accordingly. This complex and dynamic adjustment mechanism ensures more accurate and robust final decisions, thereby reducing the negative impact of data conflicts and improving the accuracy of the overall fusion results.

[0106] The application solves the data conflict problem in the multi-modal data fusion process by acquiring multi-modal data fusion parameters in real time and performing dynamic monitoring, thereby improving the accuracy and robustness of the risk assessment of latent tuberculosis infection transforming into active tuberculosis. Specifically, the key features reflecting data conflicts are extracted through feature engineering technology, and the conflicts between data sources are identified in time by means of a pre-learned deep learning model for intelligent assessment. After detecting the conflicts, the weights of each data source are dynamically adjusted based on the reliability and conflict severity of each data source, so as to reduce the negative impact of the conflict data source on the final decision. This scheme not only optimizes the precision of the multi-modal data fusion process and avoids the risk of misdiagnosis and missed diagnosis, but also improves the adaptability of the model in a complex clinical data environment, ultimately promoting the implementation of precision medicine and personalized treatment.

[0107] The above formulas are dimensionless values calculated, the formulas are obtained by collecting a large amount of data to simulate a formula of the latest real situation, and the preset parameters in the formulas are set by a person skilled in the art according to the actual situation.

[0108] The above only describes certain exemplary embodiments of the application by way of illustration, and it is self-evident that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the application. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the application.

[0109] It should be noted that in this paper, if there are relationship terms such as first and second, they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. Without more limitations, the element defined by the sentence "includes a" does not exclude the presence of other identical elements in the process, method, article or device including the element.

[0110] It should be understood that in various embodiments of the present application, the size of the sequence number of each process does not mean the order of execution, and the execution order of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0111] Those skilled in the art can clearly understand that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0112] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0113] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, i.e. they can be located in one place or distributed on multiple network units. Part or all of the units can be selected to achieve the purpose of the embodiment according to actual needs.

[0114] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can be physically present separately, or two or more units can be integrated into one unit.

[0115] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any skilled person in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0116] The above only describes some exemplary embodiments of the present application by way of illustration, and it is self-evident that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present application. Therefore, the above figures and descriptions are illustrative in nature and should not be understood as limiting the scope of protection of the claims of the present application.

Claims

1. A deep learning assisted prevention and treatment management platform for latent tuberculosis infection, characterized in that, The conflict detection method comprises a data acquisition and dynamic monitoring module, a data preprocessing and organization module, a conflict feature extraction and analysis module, a deep learning conflict evaluation module, and a dynamic weight adjustment and optimization module. The data acquisition and dynamic monitoring module acquires fusion parameter information of multi-modal data in real time, ensuring dynamic monitoring of each input during data fusion. The data preprocessing and organization module organizes the acquired multi-modal data fusion parameter information into a unified data set and pre-processes the multi-modal data fusion parameter information in the data set. The conflict feature extraction and analysis module extracts key features reflecting conflicts in data fusion from the pre-processed data through feature engineering techniques and analyzes the extracted key features. The deep learning conflict evaluation module inputs the analyzed key features into a pre-learned deep learning model to identify whether there is a conflict in the data fusion process. The dynamic weight adjustment and optimization module adjusts the weights dynamically according to the reliability of each data source and the severity of the conflict after detecting the conflict in the data fusion process. The key features reflecting conflicts in data fusion are extracted from the pre-processed data through feature engineering techniques. The extracted features include the contribution change of each data source to the final decision in the multi-modal fusion process and the error change in the multi-modal data fusion process. The contribution change of each data source to the final decision in the multi-modal fusion process and the error change in the multi-modal data fusion process are analyzed in the detection window to generate data source contribution change reference values and data fusion error reference values, respectively. The data source contribution change reference values quantify the influence and contribution change of each data source in the multi-modal data fusion process, and the data fusion error reference values measure the inconsistency in the multi-modal data fusion process over time. The specific steps for analyzing the contribution change of each data source to the final decision in the multi-modal fusion process to generate data source contribution change reference values are as follows: In the process of multi-modal data fusion, the contribution of each data source to the final decision will be constantly adjusted with the change of data input. To quantify the contribution of data sources at each time point, a dynamic weight function is introduced, which is calculated by considering the feature importance of data sources and the response of the fusion model. The calculation expression is as follows: , wherein, is the contribution weight of the jth data source to the decision at time t, is the contribution weight of the jth data source to the decision at time t, is the feature intensity of the jth data source at time t, is the feature intensity of the jth data source at time t, is the dynamic adjustment factor of the jth data source at time t, is the feature intensity of the jth data source at time t, is the dynamic adjustment factor of the jth data source at time t, is the feature intensity of the jth data source at time t, is the dynamic adjustment factor of the jth data source at time t, is the feature intensity of the jth data source at time t, is the dynamic adjustment factor of the jth data source at time t, is the feature intensity of the jth data source at time t, is the dynamic adjustment factor of the jth data source at time t, is the feature intensity of the jth data source at time t, is the dynamic adjustment factor of the jth data source at time t, is the total number of all data sources used in the multi-modal data fusion process. Next, by time series analysis on the dynamic contribution of each data source, the change of contribution of each data source is calculated, the contribution change index is introduced, the contribution change of data source between the current time and the previous time is quantified, and the calculation expression is as follows: , In the formula, It is the contribution change index, representing the first Data sources at time The contribution change index It is the first Data sources in time The weight of each time's contribution to the decision-making process. It is the first Conflict adjustment factor for each data source, It is a time-dependent factor, representing the first... The impact of external environmental changes on a data source at the current moment; Finally, the conflict of all data sources in the multi-modal data fusion is evaluated as a whole, and the contribution change index of each data source is calculated The fusion is performed to generate a data source contribution change reference value, and the calculation expression is as follows: , In the formula, is the data source contribution change reference value; The specific steps for analyzing the error change in the multi-modal data fusion process to generate data fusion error reference values are as follows: In the first step of generating the data fusion error index, the error of each data source is first calculated, and the error of each data source represents the deviation of the data source to the final prediction result. The error calculation expression is as follows: , wherein, is the error of the th data source, is the prediction result of the th data source on the th sample, is the actual value of the th sample, is the error amplification index of the th data source, is the total number of samples in the dataset, i.e., the number of samples processed when the model makes a prediction. After calculating the error of each data source, the obtained errors are weighted and fused to obtain a comprehensive error value. The error of each data source is weighted according to its importance to the fusion result, and the calculation expression is as follows: , wherein, is the total error after fusion of all data sources, is the weight of the th data source, is the error amplification index of the th data source, is the total number of all data sources used in the multi-modal data fusion process; Finally, the fused error value is converted into a data fusion error reference value for quantifying the conflict degree in the multi-modal data fusion process. A nonlinear function is used to map the fused error, and the calculation expression is as follows: , wherein is a data fusion error reference value, is a weight coefficient of the th data source, is a weighted non-linear transformation of the th data source error.

2. The platform according to claim 1, wherein The specific steps for organizing the acquired multi-modal data fusion parameter information into a unified data set are as follows: first, label and normalize the fusion parameters of each data source; Next, denoise, handle missing values, and detect anomalies in the data; Subsequently, adjust the data to a uniform scale through standardization techniques; Finally, the pre-processed data set is used for further feature extraction, conflict detection, and input to the deep learning model.

3. The tuberculosis latent infection deep learning assisted prevention and treatment management platform of claim 1, wherein, The analyzed data source contribution change reference value and data fusion error reference value are input into the pre-learned deep learning model, a fusion conflict risk coefficient is generated through the deep learning model, and the multi-modal data fusion process is intelligently evaluated through the fusion conflict risk coefficient to identify whether there is a conflict in the data fusion process.

4. The tuberculosis latent infection deep learning assisted prevention and treatment management platform according to claim 3, wherein, The fusion conflict risk coefficient generated when the multi-modal data fusion process is intelligently evaluated through the pre-learned deep learning model to identify whether there is a conflict in the data fusion process is compared and analyzed with the pre-set fusion conflict risk coefficient, the multi-modal data fusion process is divided, and the division steps are as follows: If the fusion conflict risk coefficient is greater than the pre-set fusion conflict risk coefficient reference threshold, the current multi-modal data fusion process is divided into a data fusion conflict; If the fusion conflict risk coefficient is less than or equal to the pre-set fusion conflict risk coefficient reference threshold, the current multi-modal data fusion process is divided into a data fusion conflict.

5. The tuberculosis latent infection deep learning assisted prevention and treatment management platform according to claim 4, wherein, After detecting the conflict in the data fusion process and evaluating its severity, the specific steps of dynamically adjusting the weight according to the reliability of each data source and the severity of the conflict are as follows: After detecting the conflict in the data fusion process, the severity of the conflict needs to be evaluated first, and the reliability of each data source is quantitatively evaluated, and the evaluation formula is as follows: ， wherein, is a conflict severity index of the data source representing the severity of the conflict generated by the data source in the process of multi-modal data fusion, is a conflict sensitivity parameter of the data source adjusting the degree of its reaction to the conflict, is a fusion conflict risk coefficient, is a fusion conflict risk coefficient reference threshold, is an exponential growth factor of the data source ; In the assessment of the reliability of the data source, combined with multiple weighting standardization, and joined the weight adjustment of historical performance, while considering the weighted influence of data quality and historical performance, and through the composite function to balance the importance, the calculation expression is as follows: , wherein, is the reliability score of the data source , is the quality score of the data source , is the historical performance score of the data source , is the stability coefficient of each data source, reflecting the weighted influence of historical performance on reliability assessment, is the historical variability of the data source , indicating the stability of the data source in history, is the mean of the historical variability of all data sources, and are the weight coefficients of the quality score of the data source and the historical performance score of the data source , and , respectively. In the final dynamic weight adjustment, in addition to adjusting the conflict severity and reliability, a composite weighting adjustment function is introduced, an exponential weighting is added in the adjustment process, the interaction between the conflict severity and the data reliability is considered, and an enhanced contrast adjustment is introduced. According to the conflict severity and the reliability evaluation results of the data sources, the weight of each data source is dynamically adjusted, and the calculation formula is as follows: , In the formula, It is the adjusted number The weight of each data source, These are the original weights. It is an adjustment coefficient that controls the data source. Adjusting sensitivity, These are non-linear weighting coefficients for reliability scoring. This is a conflict severity adjustment factor that controls the impact of conflict on weight adjustments. It is a weighted index of conflict severity, which modulates the nonlinear effect of conflict severity on the weight.

Images