Hand skin health condition detection system and method
By fusing three-dimensional structural contour data with multispectral image data, a standardized multidimensional health status vector is generated, and intelligent diagnosis is performed using a cognitive diagnostic module. This solves the problems of single information dimension and low detection efficiency in existing technologies, and achieves efficient and accurate skin health detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- THE SIXTH MEDICAL CENT OF THE CHINESE PEOPLES LIBERATION ARMY GENERAL HOSPITAL
- Filing Date
- 2026-01-13
- Publication Date
- 2026-04-21
AI Technical Summary
Existing technologies for hand skin health detection suffer from limited information dimensions, difficulty in multimodal data fusion, and insufficient intelligence in diagnostic models, resulting in low detection efficiency. Furthermore, existing scanning methods are time-consuming and have high hardware costs.
The system employs the fusion of three-dimensional structural contour data and multispectral image data. Data is acquired through an image acquisition module, and spatial registration and feature extraction are performed using a multimodal data fusion and feature extraction module to generate a standardized multidimensional health status vector. Intelligent diagnosis is then performed through a cognitive diagnosis module, and in-depth quantification is achieved by combining the comprehensive index of skin barrier function.
It achieves efficient and accurate detection of hand skin health status, significantly improves the accuracy of identifying sub-health conditions and early pathological changes, shortens detection time, reduces hardware and computing costs, and realizes intelligent and personalized detection processes.
Smart Images
Figure CN121890949A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of medical and health testing, and specifically relates to a system and method for detecting the health status of hand skin. Background Technology
[0002] Currently, the detection technology for hand skin health mainly relies on the analysis of a single information dimension or the combined measurement of multiple separate devices. For example, methods based on ordinary optical imaging can only acquire the color and two-dimensional texture of the skin surface, and cannot deeply assess the microscopic three-dimensional structure of key indicators of the skin barrier, as well as the distribution of biochemical components such as moisture and lipids. This results in limited ability to identify sub-health conditions such as dryness and roughness in the early stages. While professional skin testing instruments, such as moisture test pens and sebum testers, can provide quantitative data for specific indicators, they are usually contact-based single-point measurements. This makes it difficult to quickly and non-contactly acquire a comprehensive health profile of the entire hand, and it cannot effectively integrate measurement data of different physical dimensions to form a comprehensive diagnostic conclusion, resulting in fragmented information dimensions and the problem of "data silos." In addition, most existing scanning methods adopt a "one-size-fits-all" global high-precision scanning strategy, which not only significantly prolongs the detection time and generates a large amount of redundant data, but also increases the hardware cost and computational burden of the system. Therefore, existing technologies generally suffer from shortcomings such as limited information dimensions, difficulty in multimodal data fusion, insufficient intelligence of diagnostic models, and low detection efficiency. There is an urgent need for a comprehensive detection system and method that can efficiently and accurately integrate macroscopic skin structure and microscopic biochemical information, and perform intelligent and personalized diagnosis. Summary of the Invention
[0003] In view of the above-mentioned problems in the existing technology, the purpose of the present invention is to provide a detection system and method for the health status of hand skin, so as to improve the accuracy of current hand skin examination.
[0004] A system for detecting the health status of hand skin includes: an image acquisition module for acquiring three-dimensional structural contour data and multispectral image data of a user's hand; a multimodal data fusion and feature extraction module connected to the image acquisition module for spatially registering the three-dimensional structural contour data and multispectral image data, and extracting preliminary health features characterizing the macroscopic state of the skin; and a skin health state vector generation module connected to the multimodal data fusion and feature extraction module, the function of which is defined by the following formula, for integrating multiple extracted feature parameters into a standardized multidimensional health state vector:
[0005] ,
[0006] in, Let the health state vector be... This is a set of structural features calculated based on three-dimensional structural contour data. This is a spectral feature set calculated based on multispectral image data. is an optional thermal imaging feature set, and is a nonlinear mapping function used to map features from different modalities to a unified health state space.
[0007] The cognitive diagnosis and grading module is connected to the skin health state vector generation module. It has a built-in cognitive diagnosis model based on Bayesian network or deep convolutional neural network to receive the health state vector and infer the health status classification of the hand skin, the abnormal probability of each indicator and the corresponding confidence level.
[0008] Preferably, the image acquisition module includes:
[0009] The structured light projection unit is used to project a pre-coded structured light pattern onto the surface of the skin of the hand.
[0010] A high-speed camera unit is used to capture the deformed pattern under structured light illumination and to calculate the three-dimensional structure contour data based on the deformed pattern.
[0011] A tunable narrowband light source array is used to sequentially irradiate the skin of the hand in multiple preset wavelengths from visible light to near infrared, and the high-speed camera unit synchronously acquires the multispectral image data.
[0012] The skin health status vector generation module is used to calculate a comprehensive skin barrier function index that characterizes the integrity of the skin's physical barrier. The calculation method for the comprehensive skin barrier function index is as follows:
[0013] in,
[0014] This refers to the comprehensive index of skin barrier function. Normalized skin surface structure disorder The skin's hydration-lipid balance index, and These represent the mean and standard deviation of the skin hydration-lipid balance index in a healthy skin sample library, respectively. The formula nonlinearly quantifies the overall health level of the skin barrier by combining a structural damage penalty factor with a hydration index gain factor.
[0015] Preferably, the degree of disorder of the skin surface structure is determined by performing frequency domain transformation on the three-dimensional structural contour data to obtain its spatial frequency spectrum. In the spatial frequency spectrum, characteristic frequency regions corresponding to the ordered periodic texture of healthy skin and disordered frequency regions corresponding to disordered structures are distinguished. The disordered structures include skin groove breaks and abnormal wrinkles. The degree of disorder of the skin surface structure is quantified by calculating the proportion of the energy of the disordered frequency regions in the total frequency energy.
[0016] Preferably, the skin hydration-lipid balance index is calculated by analyzing the reflectance of multiple preset specific spectral bands in the multispectral image data. The calculation process includes selecting a first set of spectral bands sensitive to the moisture content of the stratum corneum and calculating a first normalized difference index to characterize the degree of skin hydration; simultaneously selecting a second set of spectral bands sensitive to the lipid content of the skin surface and calculating a second normalized difference index to characterize the abundance of skin lipids; and finally, the first and second normalized difference indices are differentially or nonlinearly combined to comprehensively evaluate the balance between moisture and lipids.
[0017] Preferably, the cognitive diagnosis and grading module has a built-in cognitive diagnosis model that can identify and distinguish sub-healthy states and suspected pathological states of hand skin based on the input health state vector; the sub-healthy state includes dryness, sensitivity, and roughness; the suspected pathological state includes erythema and desquamation in the early stage of contact dermatitis.
[0018] A method for detecting the health status of hand skin includes the following steps:
[0019] Step S1: Perform a quick, low-resolution scan of the user's entire hand to obtain preliminary color and temperature distribution information. Then, use a preset lightweight risk assessment algorithm to calculate and generate a preliminary health risk map of the hand skin. The map highlights areas that appear abnormal as potential risk areas.
[0020] Step S2: Based on the analysis results of the preliminary health risk map, make a decision; if the risk values shown in the map are all below the first preset threshold, the detection ends and a health conclusion is output; if there are potential risk areas with risk values higher than the first preset threshold, these areas are defined as detailed investigation target areas, and the scanning path and parameters for the detailed investigation target areas are automatically planned.
[0021] Step S3: Based on the detailed investigation strategy generated in step S2, drive the image acquisition module to acquire high-resolution three-dimensional structure and multispectral data only for the detailed investigation target area.
[0022] Step S4: Integrate macro background information and precise data of the target area to extract and construct a comprehensive health status vector. Input the health status vector into a pre-trained cognitive diagnostic model for comprehensive evaluation, and finally generate a personalized health profile report that includes global assessment and local lesion analysis.
[0023] Preferably, in step S1, the rapid, low-resolution scan includes: using a wide-angle camera to capture a color image of the hand, and calculating color non-uniformity in a preset color space that separates the luminance component and the chromaticity component, for preliminary screening of pigmentation or erythema areas; using an infrared thermal imaging sensor to capture a temperature distribution map of the hand, and identifying abnormal temperature points caused by local inflammation or abnormal blood circulation as potential risk areas.
[0024] Preferably, in step S2, the decision-making logic is as follows: when the risk value of any pixel in the risk map exceeds a first preset threshold, the pixel and its neighborhood are marked as a detailed investigation target area; when the average risk value in any detailed investigation target area further exceeds a second preset threshold, the system will automatically add additional characteristic spectral bands to the area for detailed investigation, for more in-depth pathological analysis.
[0025] In step S4, the fusion diagnosis employs a hierarchical diagnostic model, which includes: a local lesion analysis layer, used to quantify and classify the microscopic pathological features of the target area based on the high-resolution detailed examination data obtained in step S3, the microscopic pathological features including the severity of desquamation and the density of vesicles; and a global correlation inference layer, used to receive the classification results of the local lesion analysis layer and, in conjunction with the global macroscopic information obtained in step S1, to comprehensively infer the cause and development trend of the lesions, the global macroscopic information including overall skin color and the extent of dryness.
[0026] Compared to existing technologies, the advantages of this invention are as follows: By deeply fusing three-dimensional structural contour data with multispectral image data, this invention constructs a standardized multidimensional health state vector, thereby overcoming the shortcomings of existing technologies, such as single information dimension and fragmented data. This multimodal fusion mechanism can not only comprehensively evaluate from both physical structure and biochemical component levels, but also deeply quantify skin health through innovative parameters such as the comprehensive skin barrier function index, significantly improving the accuracy and reliability of identifying sub-health states such as dryness and sensitivity, as well as early pathological changes.
[0027] This invention first performs a rapid, low-resolution macroscopic pre-scan to identify potential risk areas, and then performs a high-precision multimodal detailed examination only on these risk areas. This strategy abandons the inefficient global scanning mode of traditional detection methods, greatly shortens the detection time, reduces data redundancy, and lowers system hardware and computing costs. It realizes intelligent and personalized detection processes, enabling diagnostic resources to be precisely focused on key lesion areas, thereby significantly improving overall detection efficiency while ensuring diagnostic accuracy. Attached Figure Description
[0028] Figure 1 This is a schematic diagram of the module structure of the detection system of the present invention.
[0029] Figure 2 This is a schematic diagram of the image acquisition module unit structure of the present invention.
[0030] Figure 3 This is a flowchart illustrating the steps of the detection method of the present invention. Detailed Implementation
[0031] The present invention will be further described below with reference to specific embodiments.
[0032] like Figure 1 The system shown in this embodiment provides a hand skin health status detection system, including: an image acquisition module for acquiring three-dimensional structural contour data and multispectral image data of the user's hand; a multimodal data fusion and feature extraction module connected to the image acquisition module for spatially registering the three-dimensional structural contour data and multispectral image data, and extracting preliminary health features characterizing the macroscopic state of the skin; and a skin health status vector generation module connected to the multimodal data fusion and feature extraction module, whose function is defined by the following formula, for integrating the extracted multiple feature parameters into a standardized multidimensional health status vector:
[0033] ,
[0034] in, Let health status vector be used. This is a set of structural features calculated based on three-dimensional structural contour data. This is a spectral feature set calculated based on multispectral image data. is an optional thermal imaging feature set, and is a nonlinear mapping function used to map features from different modalities to a unified health state space.
[0035] In one specific embodiment, the three-dimensional structural contour data is high-precision surface topography data in the form of point clouds or depth maps, with a spatial resolution of up to 50 micrometers, capable of clearly characterizing the microstructure of skin grooves and canaliculi. The multispectral image data includes reflectance images in multiple narrowband bands from visible light to near-infrared. Spatial registration is achieved through pre-calibrated camera intrinsic and extrinsic parameters and the transformation matrix between the structured light system and the multispectral imaging system, ensuring that each three-dimensional spatial point can accurately correspond to its pixel value in different spectral images. The function f can be a pre-trained deep neural network, such as a multilayer perceptron (MLP), which receives the stitched data. The features are taken as input, and the output is a vector V with fixed dimensions that can comprehensively represent the skin's health status.
[0036] The cognitive diagnosis and grading module, connected to the skin health status vector generation module, incorporates a cognitive diagnosis model based on a Bayesian network or deep convolutional neural network. This model receives the health status vector and infers the health status classification of the hand skin, the probability of abnormality for each indicator, and the corresponding confidence level. For example, this cognitive diagnosis model is a deep convolutional neural network based on the ResNet architecture, and its input is a feature map reconstructed from the health status vector V. The model is trained on a large amount of clinical data annotated by dermatologists and can output multi-dimensional diagnostic results. For example, for an input vector V, the model might output: "Sub-health state - dryness: probability 95%, confidence level 0.98", "Sub-health state - roughness: probability 88%, confidence level 0.95", "Suspected pathological state - early erythema of contact dermatitis: probability 60%, confidence level 0.75". Such outputs provide users with a quantitative, multi-level assessment of their health status.
[0037] like Figure 2 The image acquisition module shown in this embodiment includes:
[0038] A structured light projection unit is used to project a pre-coded structured light pattern onto the skin surface of the hand. In this embodiment, the structured light projection unit is a digital light processing projector capable of projecting a series of optimized sinusoidal stripe patterns or Gray code patterns at high speed. These patterns are designed to maximize the accuracy of 3D reconstruction and are robust to different curvatures and reflectivities of the skin surface.
[0039] A high-speed camera unit is used to capture the deformed patterns under structured light illumination and to calculate the three-dimensional structural contour data based on the deformed patterns.
[0040] A tunable narrowband light source array is used to sequentially irradiate the skin of the hand in multiple preset wavelengths from visible light to near infrared, and multispectral image data is simultaneously acquired by a high-speed camera unit.
[0041] The skin health status vector generation module is used to calculate a comprehensive skin barrier function index that characterizes the integrity of the skin's physical barrier. The calculation method for the comprehensive skin barrier function index is as follows:
[0042] ,in,
[0043] It is a comprehensive index of skin barrier function. Normalized skin surface structure disorder The skin's hydration-lipid balance index, and These represent the mean and standard deviation of the skin hydration-lipid balance index (CSBFI) in a healthy skin sample library. The formula nonlinearly quantifies the overall health level of the skin barrier by combining a structural damage penalty factor with a hydration index gain factor. Specifically, the CSBFI value is designed to range from 0 to 2. When the skin structure is intact (Snorm close to 0) and the hydration-lipid balance is ideal (Hindex close to μH), the CSBFI value is close to 1, representing a healthy skin barrier. When the barrier is damaged, such as with skin groove rupture (increased Snorm), the CSBFI will decrease significantly. When the skin's hydration is far better than average (Hindex much greater than μH), the index gain factor will make the CSBFI value greater than 1, indicating that the skin is in a very moisturized state. and The values are based on a database of over 5,000 healthy hand skin samples of different ages, genders, and skin types, ensuring the universality and accuracy of the assessment.
[0044] The degree of disorder in skin surface structure is determined by performing frequency domain transformation on the three-dimensional structural contour data to obtain its spatial frequency spectrum. In the spatial frequency spectrum, characteristic frequency regions corresponding to the ordered periodic texture of healthy skin and disordered frequency regions corresponding to disordered structures are distinguished. Disordered structures include skin groove breaks and abnormal wrinkles. The degree of disorder in skin surface structure is quantified by calculating the proportion of energy in disordered frequency regions in the total frequency energy.
[0045] The skin hydration-lipid balance index is calculated by analyzing the reflectance of multiple preset specific spectral bands in multispectral image data. The calculation process includes selecting a first set of spectral bands that are sensitive to the moisture content of the stratum corneum and calculating a first normalized difference index to characterize the skin hydration level. At the same time, a second set of spectral bands that are sensitive to the lipid content of the skin surface are selected and a second normalized difference index is calculated to characterize the abundance of skin lipids. Finally, the first and second normalized difference indices are differentially or nonlinearly combined to comprehensively assess the balance between moisture and lipids.
[0046] For example, the first set of spectral bands selects the 940nm band, which is sensitive to water absorption, and the 850nm band as a reference, to calculate the Normalized Differential Water Index (NDWI) = (R850 - R940) / (R850 + R940) to characterize the degree of hydration. The second set of spectral bands may utilize the absorption characteristics of lipids in specific infrared bands for calculation. Finally, the Hindex can be comprehensively evaluated using a weighted formula Hindex = w1 × NDWI - w2 × NDLI + b, where NDLI is the Normalized Differential Lipid Index, w1 and w2 are weighting coefficients, and b is a bias constant. These parameters are calibrated through regression analysis with the measurement results of a standard skin testing instrument.
[0047] The cognitive diagnosis and grading module, with its built-in cognitive diagnosis model, can identify and differentiate between sub-healthy and suspected pathological states of hand skin based on the input health status vector. Sub-healthy states include dryness, sensitivity, and roughness; suspected pathological states include early-stage erythema and desquamation of contact dermatitis. For example, when the health status vector V received by the model shows low values for the feature component related to CSBFI while abnormally high values for the feature component related to the spectral red channel (reflecting hemoglobin), the model will infer a diagnosis of "suspected contact dermatitis" and provide a description of the signs of "erythema." Simultaneously, the system will suggest that the user pay attention to changes in this area and consult a professional doctor if the abnormality persists, thus achieving an early warning function.
[0048] like Figure 3 The following is a method for detecting the health status of hand skin provided in this embodiment, including the following steps:
[0049] Step S1 involves a rapid, low-resolution scan of the user's entire hand to obtain preliminary color and temperature distribution information. A pre-defined lightweight risk assessment algorithm then generates a preliminary health risk map of the hand's skin, highlighting potentially abnormal areas as potential risk zones. In S1, the user places their hand in a designated area of the device, and a wide-angle camera captures a color image of the hand under standard D65 lighting. Simultaneously, an infrared thermal imaging sensor captures a temperature distribution map. The risk assessment algorithm converts the color image to the Lab color space, calculates local color uniformity, and combines this with anomalies exceeding the average skin temperature by 1.5°C in the temperature map to generate a pixel-level risk score (between 0 and 1), which is ultimately visualized as a risk map. The entire process takes less than one second.
[0050] Step S2: Based on the analysis results of the preliminary health risk map, a decision is made. If all risk values shown in the map are below a first preset threshold, the detection ends and a health conclusion is output. If there are potential risk areas with risk values higher than the first preset threshold, these areas are defined as detailed investigation target areas, and the scanning path and parameters for the detailed investigation target areas are automatically planned. For example, the first preset threshold is set to 0.6. The system will traverse all pixels in the risk map. Once a pixel with a risk value greater than 0.6 is found, it and its surrounding 5mm x 5mm neighborhood are marked as a detailed investigation target area. If there are multiple discontinuous risk areas, the system will plan an optimal scanning path, for example, driving the high-resolution acquisition module to conduct detailed investigations of these areas sequentially in the order from near to far in the most efficient way.
[0051] Step S3: Based on the detailed investigation strategy generated in step S2, the image acquisition module is driven to acquire high-resolution 3D structural and multispectral data only for the target area to be investigated. The drive command controls a 2D precision moving platform or galvanometer system to precisely align the field of view center of the high-resolution image acquisition module with the center point of each target area planned in S2, and triggers a high-resolution scan. This "targeted" scanning mode avoids time-consuming high-precision scanning of the entire hand, reducing the detection time from several minutes to less than 30 seconds, and greatly reducing the burden of data processing.
[0052] Step S4: Integrate macro background information and precise data of the target area to extract and construct a comprehensive health status vector. Input the health status vector into a pre-trained cognitive diagnostic model for comprehensive evaluation, and finally generate a personalized health profile report that includes global assessment and local lesion analysis.
[0053] In step S1, the fast, low-resolution scan includes: using a wide-angle camera to capture a color image of the hand and calculating color non-uniformity in a preset color space that separates the luminance component from the chromaticity component, for preliminary screening of pigmentation or erythema areas; using an infrared thermal imaging sensor to capture a temperature distribution map of the hand and identify abnormal temperature points caused by local inflammation or abnormal blood circulation as potential risk areas.
[0054] In step S2, the decision-making logic is as follows: when the risk value of any pixel in the risk map exceeds the first preset threshold, the pixel and its neighborhood are marked as a detailed investigation target area; when the average risk value in any detailed investigation target area further exceeds the second preset threshold, the system will automatically add additional characteristic spectral bands to the area for detailed investigation, in order to conduct more in-depth pathological analysis.
[0055] In step S4, the fusion diagnosis adopts a hierarchical diagnostic model, which includes: a local lesion analysis layer, used to quantify and classify the microscopic pathological features of the target area based on the high-resolution detailed examination data obtained in step S3. The microscopic pathological features include the severity of desquamation and the density of vesicles; and a global correlation inference layer, used to receive the classification results of the local lesion analysis layer and, in combination with the global macroscopic information obtained in step S1, to make a comprehensive inference on the cause and development trend of the lesions. The global macroscopic information includes the overall skin color and the extent of dryness.
[0056] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.
Claims
1. A system for detecting the health status of hand skin, characterized in that, include: An image acquisition module is used to acquire three-dimensional structural contour data and multispectral image data of the user's hand; a multimodal data fusion and feature extraction module is connected to the image acquisition module and is used to spatially register the three-dimensional structural contour data and multispectral image data, and extract preliminary health features that characterize the macroscopic state of the skin. The skin health status vector generation module, connected to the multimodal data fusion and feature extraction module, is functionally defined by the following formula, used to integrate the extracted multiple feature parameters into a standardized multidimensional health status vector: ， in, Let the health state vector be... This is a set of structural features calculated based on three-dimensional structural contour data. This is a spectral feature set calculated based on multispectral image data. is an optional thermal imaging feature set, and is a nonlinear mapping function used to map features from different modalities to a unified health state space; The cognitive diagnosis and grading module is connected to the skin health status vector generation module. It has a built-in cognitive diagnosis model based on Bayesian network or deep convolutional neural network to receive the health status vector and infer the health status classification of the hand skin, the abnormal probability of each indicator and the corresponding confidence level.
2. The hand skin health detection system according to claim 1, characterized in that: The image acquisition module includes: A structured light projection unit is used to project a pre-coded structured light pattern onto the surface of the skin of the hand; A high-speed camera unit is used to capture the deformed pattern under structured light illumination and to calculate the three-dimensional structure contour data based on the deformed pattern. A tunable narrowband light source array is used to sequentially irradiate the skin of the hand in multiple preset wavelengths from visible light to near infrared, and the high-speed camera unit synchronously acquires the multispectral image data.
3. The hand skin health detection system according to claim 1, characterized in that: The skin health status vector generation module is used to calculate a comprehensive skin barrier function index that characterizes the integrity of the skin's physical barrier. The calculation method for the comprehensive skin barrier function index is as follows: in, This refers to the comprehensive index of skin barrier function. Normalized skin surface structure disorder The skin's hydration-lipid balance index, and These represent the mean and standard deviation of the skin hydration-lipid balance index in a healthy skin sample library, respectively. The formula nonlinearly quantifies the overall health level of the skin barrier by combining a structural damage penalty factor with a hydration index gain factor.
4. The hand skin health detection system according to claim 3, characterized in that: The degree of disorder in the skin surface structure is determined by performing a frequency domain transformation on the three-dimensional structural contour data to obtain its spatial frequency spectrum. In the spatial frequency spectrum, characteristic frequency regions corresponding to the ordered periodic texture of healthy skin and disordered frequency regions corresponding to disordered structures are distinguished. The disordered structures include skin groove breaks and abnormal wrinkles. The degree of disorder in the skin surface structure is quantified by calculating the proportion of the energy of the disordered frequency regions in the total frequency energy.
5. The hand skin health detection system according to claim 3, characterized in that: The skin hydration-lipid balance index is calculated by analyzing the reflectance of multiple preset specific spectral bands in the multispectral image data. The calculation process includes selecting a first set of spectral bands that are sensitive to the moisture content of the stratum corneum and calculating a first normalized difference index to characterize the skin hydration level; simultaneously selecting a second set of spectral bands that are sensitive to the lipid content of the skin surface and calculating a second normalized difference index to characterize the abundance of skin lipids; and finally, the first and second normalized difference indices are differentially or nonlinearly combined to comprehensively evaluate the balance between moisture and lipids.
6. The hand skin health detection system according to claim 1, characterized in that: The cognitive diagnosis and grading module has a built-in cognitive diagnosis model that can identify and distinguish sub-healthy states and suspected pathological states of hand skin based on the input health state vector. The sub-healthy states include dryness, sensitivity, and roughness. The suspected pathological states include erythema and desquamation in the early stages of contact dermatitis.
7. A method for detecting the health status of hand skin, characterized in that, Includes the following steps: Step S1: Perform a quick, low-resolution scan of the user's entire hand to obtain preliminary color and temperature distribution information. Then, use a preset lightweight risk assessment algorithm to calculate and generate a preliminary health risk map of the hand skin. The map highlights areas that appear abnormal as potential risk areas. Step S2: Make a decision based on the analysis results of the preliminary health risk profile; If all risk values shown in the graph are below the first preset threshold, the detection ends and a health conclusion is output; if there are potential risk areas with risk values higher than the first preset threshold, these areas are defined as target areas for detailed investigation, and the scanning path and parameters for the target areas for detailed investigation are automatically planned. Step S3: Based on the detailed investigation strategy generated in step S2, drive the image acquisition module to acquire high-resolution three-dimensional structure and multispectral data only for the detailed investigation target area; Step S4: Integrate macro background information and precise data of the target area to extract and construct a comprehensive health status vector. Input the health status vector into a pre-trained cognitive diagnostic model for comprehensive evaluation, and finally generate a personalized health profile report that includes global assessment and local lesion analysis.
8. The method for detecting the health status of hand skin according to claim 7, characterized in that: In step S1, the rapid, low-resolution scan includes: using a wide-angle camera to capture a color image of the hand, and calculating color non-uniformity in a preset color space that separates the luminance component and the chromaticity component, for preliminary screening of pigmentation or erythema areas; using an infrared thermal imaging sensor to capture a temperature distribution map of the hand, and identifying abnormal temperature points caused by local inflammation or abnormal blood circulation as potential risk areas.
9. A method for detecting the health status of hand skin according to claim 7, characterized in that: In step S2, the decision-making logic is as follows: when the risk value of any pixel in the risk map exceeds the first preset threshold, the pixel and its neighborhood are marked as a detailed investigation target area; when the average risk value in any detailed investigation target area further exceeds the second preset threshold, the system will automatically add additional characteristic spectral bands to the area for detailed investigation, in order to conduct more in-depth pathological analysis.
10. A method for detecting the health status of hand skin according to claim 7, characterized in that: In step S4, the fusion diagnosis employs a hierarchical diagnostic model, which includes: a local lesion analysis layer, used to quantify and classify the microscopic pathological features of the target area based on the high-resolution detailed examination data obtained in step S3, the microscopic pathological features including the severity of desquamation and the density of vesicles; and a global correlation inference layer, used to receive the classification results of the local lesion analysis layer and, in conjunction with the global macroscopic information obtained in step S1, to comprehensively infer the cause and development trend of the lesions, the global macroscopic information including overall skin color and the extent of dryness.