An intelligent physical examination analysis system with a brain disease risk map construction function

Abstract

This invention discloses an intelligent health checkup and analysis system with brain disease risk map construction capabilities. The system comprises: a multimodal data acquisition and standardization module, a brain structure / functional partitioning mapping engine, a brain disease risk map construction module, a dynamic tracking and evolution prediction module, and an interactive visualization and report generation module. The multimodal data acquisition and standardization module integrates and standardizes blood indicators, cognitive behavioral data, wearable physiological data, and imaging data. The brain structure / functional partitioning mapping engine generates risk coefficients. The brain disease risk map construction module generates a three-dimensional color-coded risk map. The dynamic tracking and evolution prediction module predicts changes over the next 1-5 years, enabling dynamic tracking and early warning. The interactive visualization and report generation module displays risk data using three-dimensional heatmaps and evolution animations, reducing reliance on specialist doctors and improving the accessibility and operability of brain health management.

Description

Technical Field

[0001] This invention relates to the field of medical examination technology, specifically to an intelligent medical examination analysis system with the function of constructing a brain disease risk map. Background Technology

[0002] Brain diseases, including neurodegenerative diseases, cerebrovascular diseases, and inflammatory demyelinating diseases, have become one of the most disabling and fatal disease categories worldwide. Taking Alzheimer's disease as an example, the prevalence rate in people over 65 years old in my country is about 5% to 7%, and it is rising rapidly with the aging of the population. The existing physical examination system's brain health assessment methods include neuropsychological scales, single imaging examinations, and traditional health risk assessments, which have the following problems: Data silos: Blood biochemistry, cognitive scales, genotyping, and imaging examinations belong to different departments and lack a unified analytical framework; The risk expression is crude: it can only output qualitative conclusions of "high risk / low risk", and cannot reflect the spatial heterogeneity of risk in different anatomical functional areas of the brain (such as hippocampus, basal ganglia, and prefrontal cortex); Lack of dynamic tracking: It is impossible to use historical physical examination data to construct a risk evolution trajectory, thus missing the window for early intervention; High dependence on specialist doctors: General medical examination centers and examinees have difficulty obtaining systematic and visualized risk assessments for brain diseases. Summary of the Invention

[0003] The purpose of this invention is to provide an intelligent physical examination analysis system with brain disease risk map construction function, which solves the problem that existing physical examination systems cannot perform multimodal data fusion, spatial modeling, dynamic prediction and intuitive visualization of brain disease risk.

[0004] The present invention solves the above-mentioned technical problems through the following technical solutions, the present invention comprising: A multimodal data acquisition and standardization module is used to acquire the examinee's blood indicators, cognitive behavioral data, wearable physiological data, and optional imaging data. A brain structure / functional partitioning mapping engine is used to map standardized data to regions of interest in a standard brain atlas. and generate each risk coefficient ; A brain disease risk map construction module, which is used to construct a brain disease risk map based on the risk coefficient vector. Calculate each Based on the incidence probability of at least one brain disease, a three-dimensional color-coded risk map is generated; The dynamic tracking and evolution prediction module is used to store the risk coefficient sequence of each physical examination and predict future risk changes. An interactive visualization and report generation module is used to display the risk map and provide intervention suggestions.

[0005] Preferably, the brain structure / functional partitioning mapping engine uses a formula Calculate the risk coefficient ,in For the first A standardized set of physical examination characteristics For the total number of features, These are the elements of a feature-brain region weight matrix pre-trained using multimodal regression or graph convolutional networks. This refers to brain region bias. for function.

[0006] Preferably, the brain disease risk mapping module targets the first A type of brain disease, using formulas Calculate the whole-brain risk map, where Indicates the first Brain disease events, The total number of brain regions, For disease-specific intercept, brain region For disease The contribution coefficient, and through color mapping In each The conditional probability density is superimposed on the three-dimensional brain model in the form of a heatmap.

[0007] Preferably, the dynamic tracking and evolution prediction module includes a long short-term memory network. The update algorithm is as follows and ,in For the first Whole brain risk coefficient vector from a single physical examination for Hidden state For network parameters, For predicting time steps (1 to 5 years). The predicted future risk vector is used to generate an animation of risk evolution.

[0008] Preferably, the brain disease risk map construction module also supports multi-disease joint maps and employs a multi-task learning loss function. ,in For the number of disease types, For the first Number of samples for each disease For real labels, For all trainable weights, The regularization coefficient is used to enable a single risk map to simultaneously express the spatial risk distribution of multiple brain diseases through joint training.

[0009] Preferably, the brain structure / functional partitioning mapping engine employs a cross-modal transfer learning algorithm when image data is lacking. Estimate the simulated values ​​of imaging features of each brain region, among which For the predicted first Imaging features of brain regions (such as the volume of high signal in white matter and cortical thickness). Activation function The regression parameters are pre-trained based on a public image-multimodal database, and then... Substitute the aforementioned risk coefficient Calculation formula, replace or supplement the original To calculate the risk coefficient .

[0010] Preferably, in the interactive visualization and report generation module, for each brain region... Through formula Calculate the index of its main contribution characteristics for risk The corresponding feature names and values ​​are displayed on brain region labels, and personalized intervention suggestions are generated, which are then processed through decision rules. ,in This represents the predicted future risk coefficient.

[0011] Preferably, the dynamic tracking and evolution prediction module further includes a risk region clustering algorithm to automatically identify brain region clusters where the risk escalation rate exceeds a threshold. The specific algorithm is as follows: and ,in Set a preset change rate threshold (e.g., 0.15 / year). brain region The anatomical neighborhood is output as a cluster of high-risk evolutionary regions, which are then highlighted in the risk map.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: The multimodal data acquisition and standardization module integrates and standardizes blood indicators, cognitive behavioral data, wearable physiological data, and imaging data, breaking down data silos in traditional physical examinations and laying a data foundation for brain health assessment. The brain structure / functional partitioning mapping engine uses algorithms to map standardized data to regions of interest (ROIs) in brain atlases, generating risk coefficients. When imaging data is lacking, it estimates simulated values ​​of imaging features through cross-modal transfer learning, ensuring continuous and accurate assessment. The brain disease risk atlas construction module can calculate the incidence probability of ROIs for single or multiple brain diseases, generating three-dimensional color-coded risk atlases. Through multi-task learning, it visualizes the spatial distribution of risk for multiple diseases, overcoming the limitations of traditional qualitative assessment. The dynamic tracking and evolution prediction module uses Long Short-Term Memory (LSTM) networks to store historical risk sequences and predict changes over the next 1-5 years. Combined with clustering algorithms, it identifies brain regions with rapidly rising risk, enabling dynamic tracking and early warning. The interactive visualization and report generation module displays risk data with three-dimensional heatmaps and evolution animations, identifies the main contributing features of brain region risk, and generates personalized intervention suggestions, reducing reliance on specialist doctors and improving the accessibility and operability of brain health management. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the system structure in this invention. Detailed Implementation

[0014] The above-mentioned and other technical features and advantages of the present invention will be described in more detail below with reference to the accompanying drawings.

[0015] This embodiment provides a technical solution: an intelligent physical examination and analysis system with brain disease risk map construction function, such as... Figure 1 As shown, it includes a multimodal data acquisition and standardization module, a brain structure / functional partitioning mapping engine, a brain disease risk map construction module, a dynamic tracking and evolution prediction module, and an interactive visualization and report generation module; The multimodal data acquisition and standardization module is used to acquire blood indicators, cognitive and behavioral data, wearable physiological data, and optional imaging data of the examinees. The brain structure / functional partitioning engine is used to map standardized data to regions of interest in a standard brain atlas. and generate each risk coefficient ; The brain structure / functional partitioning mapping engine uses a formula. Calculate the risk coefficient ,in For the first A standardized set of physical examination characteristics For the total number of features, These are the elements of a feature-brain region weight matrix pre-trained using multimodal regression or graph convolutional networks. This refers to brain region bias. for function; The brain disease risk mapping module is used for constructing risk coefficient vectors. Calculate each Based on the incidence probability of at least one brain disease, a three-dimensional color-coded risk map is generated; The brain disease risk mapping module is specifically designed for the first A type of brain disease, using formulas Calculate the whole-brain risk map, where Indicates the first Brain disease events, The total number of brain regions, For disease-specific intercept, brain region For disease The contribution coefficient, and through color mapping In each The conditional probability density is superimposed on the three-dimensional brain model in the form of a heatmap; The dynamic tracking and evolution prediction module is used to store the risk coefficient sequence of each physical examination and predict future risk changes; The dynamic tracking and evolution prediction module includes a long short-term memory network. The update algorithm is as follows and ,in For the first Whole brain risk coefficient vector from a single physical examination for Hidden state For network parameters, For predicting time steps (1 to 5 years). The predicted future risk vector is used to generate an animation of risk evolution. The interactive visualization and report generation module is used to display risk profiles and provide intervention recommendations; Furthermore, in the interactive visualization and report generation module, for each brain region... Through formula Calculate the index of its main contribution characteristics for risk The corresponding feature names and values ​​are displayed on brain region labels, and personalized intervention suggestions are generated, which are then processed through decision-making rules. ,in This represents the predicted future risk coefficient.

[0016] Furthermore, the brain disease risk map construction module also supports joint maps for multiple diseases, employing a multi-task learning loss function. ,in For the number of disease types, For the first Number of samples for each disease For real labels, For all trainable weights, The regularization coefficient is used to enable a single risk map to simultaneously express the spatial risk distribution of multiple brain diseases through joint training.

[0017] Furthermore, the brain structure / functional partitioning mapping engine employs a cross-modal transfer learning algorithm when image data is lacking. Estimate the simulated values ​​of imaging features of each brain region, among which For the predicted first Imaging features of brain regions (such as the volume of high signal in white matter and cortical thickness). Activation function The regression parameters are pre-trained based on a public image-multimodal database, and then... Incorporating risk coefficient Calculation formula, replace or supplement the original To calculate the risk coefficient .

[0018] Furthermore, the dynamic tracking and evolution prediction module also includes a risk region clustering algorithm to automatically identify brain region clusters where the rate of risk increase exceeds a threshold. The specific algorithm is as follows: and ,in Set a preset change rate threshold (e.g., 0.15 / year). brain region The anatomical neighborhood is output as a cluster of high-risk evolutionary regions, which are then highlighted in the risk map.

[0019] Working principle: The multimodal data acquisition and standardization module integrates various data from examinees, including blood indicators (such as blood glucose, blood lipids, inflammatory markers, etc.), cognitive and behavioral data (memory, executive function scores, etc. obtained through standardized scales), wearable physiological data (such as heart rate variability, sleep structure parameters), and optional imaging data (such as MRI brain structural images or PET metabolic images). For non-imaging data, the system uses Z-score standardization or maximum-minimum normalization methods to eliminate dimensional differences. For imaging data, spatial standardization is achieved by registering with a standard brain template to ensure the consistency and comparability of data from different sources. After the data standardization is completed, the brain structure / functional area mapping engine starts working. This engine maps the standardized multimodal features to each ROI of the standard brain atlas. When the system acquires image data, it directly extracts the image features of each ROI, such as cortical thickness, gray matter volume, and functional connectivity strength, and fuses them with other modal features. If imaging data is lacking, a cross-modal transfer learning algorithm is used to predict simulated values ​​of key imaging features for each brain region using non-imaging features such as blood and cognition. For example, a pre-trained regression model is used to estimate the volume of the hippocampus or the cortical thickness of the prefrontal cortex. Subsequently, the engine... The function calculates the risk coefficient for each ROI. This coefficient comprehensively reflects the risk level of the brain region affected by various physical examination characteristics, where the feature-brain region weight matrix... It was trained using large-scale multimodal clinical data, ensuring the accurate quantification of the risk contribution of different features to specific brain regions; The brain disease risk map construction module, based on the whole-brain risk coefficient vector S, calculates the incidence probability of each ROI for each target brain disease (such as Alzheimer's disease, Parkinson's disease, and stroke). Taking Alzheimer's disease as an example, the system utilizes disease-specific intercepts. and brain region contribution coefficient The risk coefficient is determined by the logistic regression formula. Transform into conditional probability density The module uses a red-yellow-blue color mapping scheme, where red represents high risk and blue represents low risk, to overlay probability values ​​onto the surface of a three-dimensional brain model, forming an intuitive risk heatmap. For the need for joint assessment of multiple diseases, the module uses a multi-task learning loss function to jointly train multiple disease models, enabling a single risk map to simultaneously display the spatial distribution differences of different brain diseases in the brain. For example, the same brain region may present a high risk for Alzheimer's disease but a moderate risk for stroke. The dynamic tracking and evolution prediction module is responsible for the longitudinal management of physical examination data, and the system stores the risk coefficient sequence of each physical examination. Furthermore, the LSTM network is used to learn the evolution pattern of risk, and the hidden state is updated after each new physical examination data input. And based on the current risk vector Risk vectors for predicting the next 1-5 years based on historical data Meanwhile, the risk region clustering algorithm monitors the risk change rate of each brain region in real time. When the risk of a certain brain region and its anatomical neighborhood increases more rapidly than a threshold in two consecutive physical examinations. When the annual rate of change is greater than 0.15, the system automatically marks the region as a high-risk evolution area and highlights it in the 3D map to help doctors focus on brain regions where the risk is progressing rapidly. The interactive visualization and report generation module presents the above analysis results in a multi-dimensional manner. Users can observe the risk values ​​of each ROI by rotating and zooming the 3D brain model, and click on any brain region to display its main risk contribution characteristics, such as "Hippocampal risk is mainly contributed by APOEε4 genotype (+) and reduced episodic memory score (23 / 30)". The system will then display the risk values ​​based on the current risk coefficient. Main contribution characteristics and future forecast risks Personalized intervention recommendations are generated through preset decision rules. For example, for users with high risk in the hippocampus and positive apolipoprotein E, it is recommended to increase aerobic exercise three times a week, control fasting blood glucose <6.1mmol / L, and re-examine cognitive function every six months. The final report includes text descriptions, static screenshots of risk maps, and evolution animations, enabling ordinary medical examination centers and examinees to clearly understand the spatial distribution, dynamic changes, and intervention directions of brain disease risk.

[0020] The above are merely preferred embodiments of the present invention and are illustrative in nature, not restrictive. Those skilled in the art will understand that many changes, modifications, and even equivalents can be made within the spirit and scope defined by the claims of the present invention, all of which will fall within the protection scope of the present invention.

Claims

1. An intelligent physical examination and analysis system with brain disease risk map construction function, characterized in that, include: A multimodal data acquisition and standardization module is used to acquire the examinee's blood indicators, cognitive behavioral data, wearable physiological data, and optional imaging data. A brain structure / functional partitioning mapping engine is used to map standardized data to regions of interest in a standard brain atlas. and generate each risk coefficient ; A brain disease risk map construction module, which is used to construct a brain disease risk map based on the risk coefficient vector. Calculate each Based on the incidence probability of at least one brain disease, a three-dimensional color-coded risk map is generated; The dynamic tracking and evolution prediction module is used to store the risk coefficient sequence of each physical examination and predict future risk changes. An interactive visualization and report generation module is used to display the risk map and provide intervention suggestions.

2. The intelligent physical examination and analysis system with brain disease risk map construction function as described in claim 1, characterized in that, The brain structure / functional partitioning mapping engine uses a formula Calculate the risk coefficient ,in For the first A standardized set of physical examination characteristics For the total number of features, These are the elements of a feature-brain region weight matrix pre-trained using multimodal regression or graph convolutional networks. This refers to brain region bias. for function.

3. The intelligent physical examination and analysis system with brain disease risk map construction function as described in claim 2, characterized in that, The brain disease risk mapping module is aimed at the first A type of brain disease, using formulas Calculate the whole-brain risk map, where Indicates the first Brain disease events, The total number of brain regions, For disease-specific intercept, brain region For disease The contribution coefficient, and through color mapping In each The conditional probability density is superimposed on the three-dimensional brain model in the form of a heatmap.

4. The intelligent physical examination and analysis system with brain disease risk map construction function as described in claim 2, characterized in that, The dynamic tracking and evolution prediction module includes a long short-term memory network. The update algorithm is as follows and ,in For the first Whole brain risk coefficient vector from a single physical examination for Hidden state For network parameters, For predicting time steps (1 to 5 years). The predicted future risk vector is used to generate an animation of risk evolution.

5. The intelligent physical examination and analysis system with brain disease risk map construction function as described in claim 1, characterized in that, The brain disease risk map construction module also supports multi-disease joint maps and employs a multi-task learning loss function. ,in For the number of disease types, For the first Number of samples for each disease For real labels, For all trainable weights, The regularization coefficient is used to enable a single risk map to simultaneously express the spatial risk distribution of multiple brain diseases through joint training.

6. The intelligent physical examination and analysis system with brain disease risk map construction function as described in claim 2, characterized in that, The brain structure / functional partitioning mapping engine employs a cross-modal transfer learning algorithm when image data is lacking. Estimate the simulated values ​​of imaging features of each brain region, among which For the predicted first Imaging features of brain regions (such as the volume of high signal in white matter and cortical thickness). Activation function The regression parameters are pre-trained based on a public image-multimodal database, and then... Substitute the aforementioned risk coefficient Calculation formula, replace or supplement the original To calculate the risk coefficient .

7. The intelligent physical examination analysis system with brain disease risk map construction function as described in claim 4, characterized in that, In the interactive visualization and report generation module, for each brain region The index of its main contribution characteristics to risk is calculated using a formula. The corresponding feature names and values ​​are displayed on brain region labels, and personalized intervention suggestions are generated, which are then processed through decision rules. , where is the predicted future risk coefficient.

8. The intelligent physical examination and analysis system with brain disease risk map construction function as described in claim 4, characterized in that, The dynamic tracking and evolution prediction module also includes a risk region clustering algorithm to automatically identify brain region clusters where the risk escalation rate exceeds a threshold. The specific algorithm is as follows: and ,in Set a preset change rate threshold (e.g., 0.15 / year). brain region The anatomical neighborhood is output as a cluster of high-risk evolutionary regions, which are then highlighted in the risk map.

Images