Detection system for evaluating rheumatism activity

Abstract

The invention belongs to the technical field of medical health, and discloses a detection system for evaluating rheumatism activity, which comprises a data acquisition module, a multi-modal data preprocessing module, a feature fusion extraction module, an intelligent evaluation model module and a result output module which are connected in sequence. Through synchronous acquisition of multi-modal data (clinical data, image data, symptom data and behavior data), the limitation of single data dimension is broken through, the rheumatism activity can be comprehensively reflected from multiple aspects such as physiological indexes, imaging performance, subjective feeling and objective behaviors, and a foundation is laid for accurate evaluation; through combined processing of adaptive threshold filtering, normalization and time sequence interpolation, the problems of noise, dimension difference and missing of multi-modal data are effectively solved, meanwhile, time sequence features are reserved, and the accuracy of subsequent feature extraction and model operation is ensured; through combination of the deep learning network and the attention mechanism, adaptive weighted fusion of cross-modal features is realized, the complementary value of each modal data can be fully mined, the richness and distinction degree of feature expression are improved, and actual application and operation are facilitated.

Description

Technical Field

[0001] This invention relates to the field of medical and health technology, and in particular to a detection system for assessing the activity of rheumatic diseases. Background Technology

[0002] Rheumatic diseases are a class of autoimmune diseases characterized by chronic inflammation of the joints, muscles, bones and related soft tissues (such as rheumatoid arthritis, ankylosing spondylitis, systemic lupus erythematosus, etc.). Accurate assessment of disease activity (degree of inflammation activity) is a key aspect of clinical diagnosis and treatment. Accurate assessment of activity can guide the adjustment of treatment plans (such as optimization of drug dosage and type), predict the risk of irreversible damage such as joint deformities, and directly affect the patient's prognosis and quality of life.

[0003] However, existing equipment has significant limitations in clinical assessment of rheumatic disease activity. Commonly used scales such as the DAS28 (Disease Activity Scale for 28 Joints) and SDAI (Simplified Disease Activity Index) rely on physicians' subjective counting of joint swelling / tenderness and aggregated scores of a few biochemical indicators (such as erythrocyte sedimentation rate and C-reactive protein). These methods are heavily influenced by the assessor's experience and focus only on local joint symptoms, failing to reflect systemic inflammatory status and overall patient function (such as daily living abilities). Existing technologies are mostly limited to the analysis of single or a few data types, lacking a systematic integration of clinical biochemical indicators, imaging features, patient subjective symptoms, and objective behavioral data. Even some studies attempting to integrate multimodal data suffer from low information utilization due to the lack of adaptive weighting algorithms for cross-modal features (such as dynamic allocation of the contribution weights of different data to activity), making it difficult to capture the complex dynamic patterns of disease activity and hindering practical application and operation. Summary of the Invention

[0004] One object of the present invention is to provide a detection system for assessing the activity of rheumatic diseases.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: a detection system for assessing the activity of rheumatic diseases, comprising a data acquisition module, a multimodal data preprocessing module, a feature fusion and extraction module, an intelligent assessment model module, and a result output module connected in sequence;

[0006] The data acquisition module is used to simultaneously collect clinical biochemical indicators, joint images, patient-reported symptoms, and dynamic behavioral data of patients with rheumatic diseases; the multimodal data preprocessing module is used to standardize, align, and suppress noise in the collected data of various types; the feature fusion and extraction module is used to achieve adaptive weighted fusion of cross-modal features through a deep learning network; the intelligent assessment model module is used to construct a dynamic assessment model based on the fused features, and output a quantitative score and risk classification of rheumatic disease activity; the result output module is used to present the assessment results in a visual manner and provide analysis of key influencing factors.

[0007] Preferably, the data acquisition module includes a biochemical detection unit, an image acquisition unit, a symptom reporting unit, and a behavior monitoring unit; the biochemical detection unit is used to collect at least eight core inflammatory markers, including C-reactive protein, erythrocyte sedimentation rate, and anti-cyclic citrullinated peptide antibody; the image acquisition unit is used to acquire synovial thickness, blood flow signal, and bone destruction parameters in joint ultrasound images; the symptom reporting unit is used to support patients in recording information on joint pain level, swelling quantity, and fatigue index via mobile terminals; the behavior monitoring unit is used to collect data on patients' daily activity level, joint range of motion, and sleep quality via wearable devices.

[0008] Preferably, the multimodal data preprocessing module uses an adaptive threshold filtering algorithm to remove abnormal data, normalizes the dimensions of data in different dimensions through normalization, and uses time-series interpolation to fill in missing data, while retaining the time-series characteristics of the original data during the preprocessing process.

[0009] Preferably, the feature fusion extraction module includes a clinical feature extraction submodule, an image feature extraction submodule, and a temporal feature extraction submodule; the clinical feature extraction submodule uses a gradient boosting tree algorithm to screen key biochemical indicators; the image feature extraction submodule automatically identifies pathological features in joint images through a convolutional neural network; the temporal feature extraction submodule extracts the dynamic change trend of patient data based on a long short-term memory network; the outputs of each submodule are weighted and fused through an attention mechanism to generate a multi-dimensional fused feature vector.

[0010] Preferably, the intelligent assessment model module is a hybrid model based on transfer learning. It is based on the trained benchmark model for assessing the activity of rheumatic diseases, and is fine-tuned by combining the patient's individual historical data to construct a personalized assessment model. The hybrid model integrates the feature importance ranking of random forest with the nonlinear fitting ability of neural network to output a quantitative score of activity from zero to ten and a four-level risk classification of "no activity", "low activity", "medium activity" and "high activity".

[0011] Preferably, it also includes a model dynamic update module, which is used to iteratively optimize the intelligent assessment model using an incremental learning algorithm based on the assessment results of new patients and clinical follow-up data, and update the model parameters to improve the assessment adaptability to different subtypes of rheumatic diseases.

[0012] Preferably, the result output module is also used to automatically match and generate treatment plan recommendations and follow-up cycle suggestions based on the assessment results, and at the same time generate a visual report to intuitively show the contribution weight of each indicator to the activity assessment results.

[0013] Preferably, the data collected by the behavior monitoring unit includes daily walking steps, joint flexion and extension frequency, resting heart rate variability, and number of sleep interruptions at night. The data sampling frequency is once per minute, and it supports real-time synchronization with the hospital's electronic medical record system.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0015] (1) This invention breaks through the limitations of a single data dimension by synchronously collecting multimodal data (clinical, imaging, symptoms, and behavior), and can comprehensively reflect the activity of rheumatic diseases from multiple aspects such as physiological indicators, imaging manifestations, subjective feelings, and objective behaviors, laying the foundation for accurate assessment; the combination of adaptive threshold filtering, normalization, and temporal interpolation effectively solves the problems of noise, dimensional differences, and missing data in multimodal data, while preserving time series features to ensure the accuracy of subsequent feature extraction and model operation; the combination of deep learning network and attention mechanism realizes adaptive weighted fusion of cross-modal features, which can fully explore the complementary value of each modality of data and improve the richness and discriminability of feature expression.

[0016] (2) This invention, through the design of transfer learning and hybrid models (random forest + neural network), possesses both feature interpretability (feature importance of random forest) and the ability to fit complex relationships (nonlinearity of neural network); the personalized fine-tuning mechanism can also adapt to the individual differences of different patients, and the output quantitative scores and risk classifications facilitate rapid clinical decision-making; the integrated output of visual reports, treatment plan recommendations, and follow-up cycle suggestions is intuitive and practical, helping clinicians to efficiently formulate treatment strategies; the data synchronization electronic medical record system realizes the standardized management of medical data; the incremental learning-driven model dynamic update enables the system to continuously optimize with the accumulation of clinical data, continuously improve the assessment ability of different subtypes of rheumatic diseases, and realize "personalized-long-term" disease monitoring and management. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall process structure of the present invention. Detailed Implementation

[0018] The present invention will now be further described in conjunction with specific embodiments. It should be noted that, without conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments.

[0019] In the description of this invention, it should be noted that directional terms such as "center," "lateral," "longitudinal," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise" indicate the orientation and positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. They should not be construed as limiting the specific protection scope of this invention.

[0020] It should be noted that the terms "first" and "second" in the specification and claims of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0021] One preferred embodiment of the present invention, such as Figure 1 As shown, a detection system for assessing the activity of rheumatic diseases includes a data acquisition module, a multimodal data preprocessing module, a feature fusion and extraction module, an intelligent assessment model module, and a result output module connected in sequence.

[0022] The data acquisition module is used to simultaneously collect clinical biochemical indicators, joint images, patient-reported symptoms, and dynamic behavioral data of patients with rheumatic diseases; the multimodal data preprocessing module is used to standardize, align, and suppress noise in the collected data of various types; the feature fusion and extraction module is used to achieve adaptive weighted fusion of cross-modal features through a deep learning network; the intelligent assessment model module is used to construct a dynamic assessment model based on the fused features, and output a quantitative score and risk classification of rheumatic disease activity; the result output module is used to present the assessment results in a visual manner and provide analysis of key influencing factors.

[0023] The data acquisition module includes a biochemical detection unit, an image acquisition unit, a symptom reporting unit, and a behavior monitoring unit. The biochemical detection unit is used to collect at least eight core inflammatory markers, including C-reactive protein, erythrocyte sedimentation rate, and anti-cyclic citrullinated peptide antibody. The image acquisition unit is used to acquire synovial thickness, blood flow signal, and bone destruction parameters from joint ultrasound images. The symptom reporting unit is used to support patients in recording information on joint pain level, swelling quantity, and fatigue index via mobile terminals. The behavior monitoring unit is used to collect data on patients' daily activity levels, joint range of motion, and sleep quality via wearable devices.

[0024] The multimodal data preprocessing module uses an adaptive threshold filtering algorithm to remove abnormal data, normalizes the dimensions of data in different dimensions, and uses time-series interpolation to fill in missing data, while retaining the time-series characteristics of the original data during the preprocessing process.

[0025] The feature fusion and extraction module includes a clinical feature extraction submodule, an image feature extraction submodule, and a temporal feature extraction submodule. The clinical feature extraction submodule uses a gradient boosting tree algorithm to screen key biochemical indicators. The image feature extraction submodule automatically identifies pathological features in joint images through a convolutional neural network. The temporal feature extraction submodule extracts the dynamic change trend of patient data based on a long short-term memory network. The outputs of each submodule are weighted and fused through an attention mechanism to generate a multi-dimensional fused feature vector.

[0026] The intelligent assessment model module is a hybrid model based on transfer learning. It is based on the trained benchmark model for assessing rheumatic disease activity and fine-tuned by combining individual patient historical data to construct a personalized assessment model. The hybrid model integrates the feature importance ranking of random forest with the nonlinear fitting ability of neural network to output a quantitative score of activity from zero to ten and a four-level risk classification of "no activity", "low activity", "medium activity" and "high activity".

[0027] It also includes a model dynamic update module, which uses incremental learning algorithms to iteratively optimize the intelligent assessment model based on the assessment results of new patients and clinical follow-up data, and update the model parameters to improve the assessment adaptability to different subtypes of rheumatic diseases.

[0028] The results output module is also used to automatically match and generate treatment plan recommendations and follow-up cycle suggestions based on the assessment results, and at the same time generate a visual report to intuitively show the contribution weight of each indicator to the activity assessment results.

[0029] The behavioral monitoring unit collects data including daily steps, joint flexion and extension frequency, resting heart rate variability, and number of sleep interruptions at night. The data sampling frequency is once per minute, and it supports real-time synchronization with the hospital's electronic medical record system.

[0030] Working principle:

[0031] During use, clinical biochemical indicators (such as C-reactive protein, erythrocyte sedimentation rate, and other inflammatory markers), joint images (synovial thickness and blood flow signals from joint ultrasound), patient-reported symptoms (degree of joint pain, number of swollen joints, etc.), and dynamic behavioral data (daily activity levels and range of motion collected by wearable devices) of patients with rheumatic diseases are collected simultaneously to achieve multi-dimensional data coverage. An adaptive threshold filtering algorithm is used to remove abnormal data, and normalization is used to unify the dimensions of data from different dimensions. Then, time-series interpolation is used to complete the missing data while retaining the time-series characteristics of the original data, providing a high-quality, standardized dataset for subsequent analysis. Key features of each modality are extracted using gradient boosting tree algorithm (clinical features), convolutional neural network (image features), and long short-term memory network (time-series features), respectively. Then, an attention mechanism is used to adaptively weight and fuse the multi-modal features to generate a multi-dimensional fused feature vector.

[0032] A hybrid model based on transfer learning (integrating feature importance ranking in random forests with the nonlinear fitting ability of neural networks) is fine-tuned using individual patient historical data to output a quantitative activity score of 0-10 and a four-level risk classification of "no activity," "low activity," "moderate activity," and "high activity." The assessment results and the contribution weight of each indicator are presented in a visual report, while treatment plan recommendations and follow-up cycle suggestions are automatically generated, and the data is synchronized to the electronic medical record system. Based on the newly added follow-up data, the intelligent assessment model is iteratively optimized through incremental learning algorithms, and the model parameters are updated to improve the assessment adaptability to different subtypes of rheumatic diseases, enabling long-term reuse of the personalized model.

[0033] The basic principles, main features, and advantages of this invention have been described above. Those skilled in the art should understand that this invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made without departing from the spirit and scope of the invention, and all such changes and modifications fall within the scope of the invention as claimed. The scope of protection claimed by this invention is defined by the appended claims and their equivalents.

Claims

1. A detection system for assessing the activity of rheumatic diseases, characterized in that, It includes a data acquisition module, a multimodal data preprocessing module, a feature fusion and extraction module, an intelligent evaluation model module, and a result output module, which are connected in sequence. The data acquisition module is used to simultaneously collect clinical biochemical indicators, joint images, patient-reported symptoms, and dynamic behavioral data of patients with rheumatic diseases; the multimodal data preprocessing module is used to standardize, align, and suppress noise in the collected data of various types; the feature fusion and extraction module is used to achieve adaptive weighted fusion of cross-modal features through a deep learning network; the intelligent assessment model module is used to construct a dynamic assessment model based on the fused features and output a quantitative score and risk classification of rheumatic disease activity. The results output module is used to present the evaluation results in a visual manner and provide analysis of key influencing factors.

2. The detection system for assessing rheumatic disease activity as described in claim 1, characterized in that: The data acquisition module includes a biochemical detection unit, an image acquisition unit, a symptom reporting unit, and a behavior monitoring unit. The biochemical detection unit is used to collect at least eight core inflammatory markers, including C-reactive protein, erythrocyte sedimentation rate, and anti-cyclic citrullinated peptide antibody. The image acquisition unit is used to acquire synovial thickness, blood flow signal, and bone destruction parameters from joint ultrasound images. The symptom reporting unit is used to support patients in recording information on joint pain level, swelling quantity, and fatigue index via mobile terminals. The behavior monitoring unit is used to collect data on patients' daily activity levels, joint range of motion, and sleep quality via wearable devices.

3. The detection system for assessing rheumatic disease activity as described in claim 1, characterized in that: The multimodal data preprocessing module uses an adaptive threshold filtering algorithm to remove abnormal data, normalizes the dimensions of data in different dimensions, and uses time-series interpolation to fill in missing data, while retaining the time-series characteristics of the original data during the preprocessing process.

4. The detection system for assessing rheumatic disease activity as described in claim 1, characterized in that: The feature fusion and extraction module includes a clinical feature extraction submodule, an image feature extraction submodule, and a temporal feature extraction submodule. The clinical feature extraction submodule uses a gradient boosting tree algorithm to screen key biochemical indicators. The image feature extraction submodule automatically identifies pathological features in joint images through a convolutional neural network. The temporal feature extraction submodule extracts the dynamic change trend of patient data based on a long short-term memory network. The outputs of each submodule are weighted and fused through an attention mechanism to generate a multi-dimensional fused feature vector.

5. The detection system for assessing rheumatic disease activity as described in claim 1, characterized in that: The intelligent assessment model module is a hybrid model based on transfer learning. It is based on the trained benchmark model for assessing rheumatic disease activity and fine-tuned by combining individual patient historical data to construct a personalized assessment model. The hybrid model integrates the feature importance ranking of random forest with the nonlinear fitting ability of neural network to output a quantitative score of activity from zero to ten and a four-level risk classification of "no activity", "low activity", "medium activity" and "high activity".

6. The detection system for assessing rheumatic disease activity as described in claim 1, characterized in that: It also includes a model dynamic update module, which uses incremental learning algorithms to iteratively optimize the intelligent assessment model based on the assessment results of new patients and clinical follow-up data, and updates the model parameters to improve the assessment adaptability to different subtypes of rheumatic diseases.

7. The detection system for assessing rheumatic disease activity as described in claim 1, characterized in that: The results output module is also used to automatically match and generate treatment plan recommendations and follow-up cycle suggestions based on the assessment results, and at the same time generate a visual report to intuitively show the contribution weight of each indicator to the activity assessment results.

8. The detection system for assessing rheumatic disease activity as described in claim 2, characterized in that: The behavioral monitoring unit collects data including daily steps, joint flexion and extension frequency, resting heart rate variability, and number of sleep interruptions at night. The data sampling frequency is once per minute, and it supports real-time synchronization with the hospital's electronic medical record system.

Images