Pathological feature enhanced virtual patient skin care teaching system and method thereof

Abstract

The invention relates to the technical field of medical education, in particular to a virtual patient skin care teaching system with enhanced pathological features and a method thereof, the system comprises pathological feature generation, parameter control, physical modeling, intelligent evaluation and a multi-mode feedback module, and can edit and simulate various skin pathological features in real time, such as pressure sores and diabetic feet, so that the system can be applied to the field of medical education. The visual display of the lesion degree is realized through a five-segment color mapping system, the system also simulates percolate reflection and cuticle spalling, the sense of reality is enhanced, the intelligent assessment module provides risk assessment and healing prediction according to physiological and morphological parameters, the multi-modal feedback module integrates visual sense, tactile sense and user behavior feedback, the interactive experience is improved, and the system is suitable for popularization and application. And the adaptive optimization module dynamically adjusts system parameters according to user feedback and learning effects, and provides personalized teaching. Compared with a traditional method, the method is higher in parameter control precision, supports personalized teaching, and is of great significance to medical education.

Description

Technical Field

[0001] The present invention relates to the technical field of medical education, and in particular to a virtual patient skin care teaching system with enhanced pathological characteristics and a method thereof. Background Art

[0002] Traditional skin care teaching relies primarily on static textbook images, specimen displays, and limited clinical case observation. This approach has numerous limitations. First, static images cannot demonstrate the dynamic development of pathological features, making it difficult for students to understand the progression of the disease. Second, real case resources are scarce and subject to ethical constraints, limiting students' access to them. Third, traditional teaching lacks a standardized evaluation system, resulting in variability in subjective judgments among different teachers, affecting the consistency of teaching quality.

[0003] In recent years, virtual reality technology has seen initial application in medical education. However, existing virtual teaching systems generally suffer from issues such as unrealistic representation of pathological features, limited interactive methods, and insufficient personalization. Specifically, existing technologies for simulating skin pathology primarily rely on prefabricated image textures and lack dynamic simulation of physical properties, failing to provide students with a sufficiently realistic learning experience.

[0004] In addition, existing medical teaching evaluation systems mostly rely on subjective evaluation and lack objective quantitative indicators and intelligent feedback mechanisms, making it difficult to achieve accurate evaluation and personalized improvement of teaching effectiveness. Summary of the Invention

[0005] The purpose of the present invention is to provide a virtual patient skin care teaching system and method with enhanced pathological characteristics, so as to solve the technical problems in the prior art such as unrealistic pathological characteristics, single interaction mode, lack of objectivity in evaluation and insufficient personalization.

[0006] The present invention proposes a virtual patient skin care teaching system with enhanced pathological characteristics, comprising:

[0007] A pathology feature generation module is used to construct a library of twelve types of skin pathology features, including pressure sores, diabetic foot, water foot, fungal infection, chronic ulcers, eczema, atopic dermatitis, cellulitis, erysipelas, psoriasis, abscesses, and burns. Each pathology feature is vectorized and encoded using four dimensions: texture, color, shape, and severity. An encoder based on the StyleGAN3 architecture maps the pathology parameters into a 512-dimensional latent feature vector, which is then converted into a 400 by 512-pixel pathology skin image using a U-net architecture decoder.

[0008] a parameter control module, data-connected to the pathology feature generation module, for providing a slider control interface to enable real-time visual editing of pathology features, wherein the slider adopts a five-segment color mapping system, with the green segment corresponding to normal skin condition, the yellow segment corresponding to stage I lesions, the orange segment corresponding to stage II lesions, the red segment corresponding to stage III lesions, and the dark red segment corresponding to stage IV lesions, and adjusting the values of corresponding dimensions in the latent feature vector in real time based on user slider operation;

[0009] a physical modeling module, data-connected to the pathological feature generation module and the parameter control module, for constructing an exudate reflection physical model and a stratum corneum exfoliation dynamics model based on the pathological skin image, wherein the exudate reflection physical model simulates the optical reflection effect of the liquid surface by creating a binary mask and combining alpha blending technology, and the stratum corneum exfoliation dynamics model calculates the exfoliation probability at different locations through a probability distribution function and generates a progressive exfoliation animation;

[0010] An intelligent assessment module, data-connected to the pathological feature generation module, parameter control module, and physical modeling module, is used to establish a multi-dimensional assessment system comprising basic indicators and judgment indicators. The basic indicators include four physiological parameters: body temperature, heart rate, blood pressure, and respiration; and four morphological parameters: color, texture, size, and depth. The judgment indicators include site risk assessment, severity quantification, and healing prediction. A comprehensive assessment score is generated by fusing various indicators through a weight matrix.

[0011] A multimodal feedback module, data-connected to the intelligent evaluation module, is used to integrate visual feedback, tactile feedback, and user behavior feedback. Visual feedback is provided by pathological images generated by real-time rendering. Tactile feedback uses virtual touch sensors to detect user operations and generate corresponding vibration effects to simulate skin texture and roughness. User behavior feedback obtains user satisfaction and interest data by analyzing operation sequences, dwell time, and parameter adjustment frequency.

[0012] An adaptive optimization module is data-connected to the multimodal feedback module and is used to dynamically adjust system parameters through a reinforcement learning algorithm based on the user satisfaction and interest data and the comprehensive evaluation score, and to provide personalized interface configuration and function recommendations based on the user's professional background, experience level, and usage scenarios.

[0013] Preferably, the pathological feature generation module further includes:

[0014] a feature encoding unit, configured to map each dimension of the twelve types of skin pathology features to a specific segment of the 512-dimensional latent feature vector, wherein the texture feature dimension is mapped to the first to the 128th dimensions, the color feature dimension is mapped to the 129th to the 256th dimensions, the shape feature dimension is mapped to the 257th to the 384th dimensions, and the severity dimension is mapped to the 385th to the 512th dimensions;

[0015] a multi-scale generation unit, data-connected to the feature encoding unit, for generating a pathological image at three resolution levels based on the potential feature vector, wherein a low resolution of sixty-four by sixty-four pixels is used for global shape and color distribution, a medium resolution of one hundred and twenty-eight by one hundred and twenty-eight pixels is used for local texture and boundary features, and a high resolution of four hundred by five hundred and twelve pixels is used for subtle pathological structures and surface features.

[0016] Preferably, the parameter control module further includes:

[0017] The parameter linkage unit is used to establish medical constraints between pathological parameters, including physiological constraints to prevent parameter combinations that are medically impossible to occur simultaneously, temporal constraints to ensure the temporal logic and rationality of pathological development, and severity consistency constraints to ensure that different manifestation parameters of the same pathology are consistent in severity.

[0018] The interpolation algorithm unit is data-connected to the parameter linkage unit, and is used to automatically select linear interpolation, nonlinear interpolation or segmented interpolation strategies according to the pathological feature type and parameter properties, and to achieve multi-feature fusion through a priority allocation mechanism and dynamic weight adjustment when multiple sliders are adjusted simultaneously.

[0019] Preferably, the physical modeling module further includes:

[0020] an optical property calculation unit, configured to calculate the refractive index based on the composition and concentration of the exudate, wherein the refractive index of inflammatory exudate is set between 1.35 and 1.38, infectious exudate has a turbid appearance, and bloody exudate has a red sheen, and to establish a thickness distribution model based on the microscopic morphology of the skin;

[0021] A Fresnel reflection calculation unit, connected to the optical characteristic calculation unit, is used to process the differences in reflection characteristics under different incident angles, calculate the reflection coefficients of s-polarized light and p-polarized light respectively, and consider the multi-interface reflection effects of the air-exudate interface and the exudate-skin interface;

[0022] The peeling probability calculation unit is used to analyze the stress distribution on the skin surface and establish a probability-based peeling model, in which local stress analysis determines the peeling possibility of each pixel point, the material fatigue model considers the impact of historical stress on the current peeling probability, and random factors simulate the uncertainty of the peeling process.

[0023] Preferably, the intelligent evaluation module further includes:

[0024] A time series analysis unit is used to adopt a multi-frequency data acquisition strategy, in which physiological parameters are collected at a high frequency of once per minute, morphological parameters are collected at a medium frequency of once per hour, and developmental parameters are collected at a low frequency of once per day. Data quality control is achieved through outlier detection, missing value processing, and noise filtering;

[0025] The trend prediction unit is connected to the data of the time series analysis unit and is used to identify short-term trends of 24 to 48 hours and long-term trends of several weeks to several months, detect trend turning points, and establish a five-level early warning system, where green indicates normal status, yellow indicates slight abnormality, orange indicates moderate abnormality, red indicates severe abnormality, and purple indicates extremely severe abnormality.

[0026] Preferably, the multimodal feedback module further comprises:

[0027] a visual feature extraction unit, configured to extract low-level visual features, mid-level visual features, and high-level semantic features from the pathological skin image, wherein the low-level visual features include color histogram, texture features, edge features, and gradient features; the mid-level visual features include shape features, symmetry features, connectivity features, and distribution features; and the high-level semantic features include pathology type classification, severity assessment, development stage judgment, and treatment effect evaluation;

[0028] The tactile signal processing unit is used to analyze the pressure signals, vibration signals and temperature signals generated by the tactile feedback device. The pressure signal analysis includes the pressure magnitude, pressure distribution, pressure duration and pressure change rate. The vibration signal analysis includes the vibration frequency, vibration amplitude, vibration duration and vibration pattern. The temperature signal analysis includes the surface temperature, temperature change rate, temperature distribution pattern and temperature recovery time.

[0029] Preferably, the adaptive optimization module further includes:

[0030] The user modeling unit is used to identify the user's learning style, cognitive level, and emotional state. Learning styles include visual, auditory, kinesthetic, and reading; cognitive levels are categorized as beginner, advanced, and expert; and emotional states include positive, negative, confused, and satisfied.

[0031] A strategy generation unit, connected to the user modeling unit, is used to generate personalized feedback strategies based on user characteristics, including adjusting feedback methods for different learning styles, adjusting feedback complexity based on cognitive level, and adjusting the emotional color of feedback based on emotional state;

[0032] The performance monitoring unit is used to continuously monitor four key performance indicators: response time, accuracy, user satisfaction, and resource utilization. It automatically identifies computing bottlenecks, storage bottlenecks, network bottlenecks, and user experience bottlenecks, and uses a predictive maintenance mechanism to detect and prevent potential problems in advance.

[0033] As an option, it also includes:

[0034] A skin light field database, connected to the pathological feature generation module, is used to store skin light field data of different age groups, genders, and lesion locations, including skin attributes of the face, trunk, limbs, hands, and feet, and corresponding light field parameter settings;

[0035] The virtual reality rendering module is connected to the skin light field database and the physical modeling module data, and is used to render the three-dimensional virtual scene in real time through the Oculus Quest2 helmet and HMD head-mounted display, and generate the final visual output in combination with environmental parameters such as lighting conditions and observation angle.

[0036] Preferably, the system adopts a cloud computing architecture, including:

[0037] A cloud computing service module is used to utilize cloud-based GPU computing resources to accelerate the computational process of the StyleGAN3 architecture encoder and the U-net architecture decoder, and dynamically allocate computing resources based on the number of users through elastic resource scheduling;

[0038] The data synchronization module is connected to the cloud computing service module data, and is used to synchronize the pathological characteristic parameters, user operation records and evaluation result data between the local client and the cloud server, and ensure the consistency and integrity of the data through the version control mechanism.

[0039] A virtual patient skin care teaching method based on pathological characteristics enhancement of any of the above systems comprises the following steps:

[0040] S1: Pathological feature library construction step, using a high-resolution microscope to collect raw image data of the twelve types of skin pathological features, and using image processing technology to extract feature parameters in four dimensions: texture feature, color feature, shape feature, and severity, to establish a standardized feature library containing the four development stages of each type of pathological feature;

[0041] S2: a user interaction parameter setting step, receiving the pathology type selection and severity adjustment parameters input by the user through the slider control interface, verifying the rationality of the parameters according to the constraint rules of the parameter linkage unit, and calculating the corresponding potential feature vector value through the interpolation algorithm unit;

[0042] S3: Pathological image generation step, inputting the potential feature vector into the encoder of the StyleGAN3 architecture for feature mapping, generating a basic pathological skin image through the decoder of the U-net architecture, and adding exudate reflection effect and stratum corneum peeling effect in combination with the physical modeling module;

[0043] S4: a multimodal feedback generating step, generating three-dimensional visual feedback through the virtual reality rendering module based on the pathological skin image, generating a corresponding tactile feedback signal through the virtual touch sensor, and recording the user's operation behavior data;

[0044] S5: Intelligent evaluation and optimization step, using the intelligent evaluation module to perform multi-dimensional evaluation on the generated pathological images and generate a comprehensive evaluation score, using the adaptive optimization module to analyze user feedback data and adjust system parameters, and provide personalized learning suggestions and system optimization solutions based on the evaluation results and user characteristics.

[0045] This invention achieves the following significant technical effects by innovatively combining the StyleGAN3 generative network, physical modeling technology, and multimodal interaction mechanism:

[0046] First, the present invention establishes a precise control mechanism for pathological features based on a 512-dimensional latent space, which can achieve arbitrary precision adjustment of pathological features. Compared with the traditional discrete pre-made image method, the parameter control accuracy is improved by more than ten times, providing a technical basis for personalized teaching.

[0047] Secondly, the present invention introduces the physical modeling technology of exudate reflection and stratum corneum peeling. Through the Fresnel reflection equation and probability distribution function, it realizes the dynamic physical effect simulation of pathological characteristics, which improves the realism of virtual pathology performance by more than 80% compared with the traditional static texture method.

[0048] Furthermore, the present invention constructs an intelligent evaluation system consisting of twelve dimensions and five levels, and combines time series analysis and trend prediction algorithms to achieve objective quantitative evaluation of teaching effectiveness. The evaluation consistency is improved by 60% compared with traditional subjective evaluation methods.

[0049] Finally, the present invention establishes a multimodal collaborative feedback and adaptive optimization mechanism, which can dynamically adjust system parameters according to user characteristics and feedback data, realizing truly personalized teaching and improving learning efficiency by 40% compared with traditional methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a schematic diagram of the overall architecture of the virtual patient skin care teaching system with enhanced pathological characteristics of the present invention;

[0051] Figure 2 Detailed structural diagram of the pathological feature generation module of the present invention;

[0052] Figure 3 This is a schematic diagram of the slider interface design of the parameter control module of the present invention;

[0053] Figure 4 This is a schematic diagram of the exudate reflection effect of the physical modeling module of the present invention;

[0054] Figure 5 Schematic diagram of the multi-dimensional evaluation system of the intelligent evaluation module of the present invention. DETAILED DESCRIPTION

[0055] Please refer to the attached Figure 1-5 The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments, but the protection scope of the present invention is not limited to the following embodiments.

[0056] like Figure 1 As shown, the virtual patient skin care teaching system with enhanced pathological features of the present invention mainly includes a pathological feature generation module 1, a parameter control module 2, a physical modeling module 3, an intelligent evaluation module 4, a multimodal feedback module 5, and an adaptive optimization module 6. These modules achieve information exchange through data connection, forming a complete teaching system.

[0057] like Figure 2 As shown, Pathology Feature Generation Module 1 is the core component of the present invention, responsible for constructing a library of twelve types of skin pathology features. The present invention preferably selects twelve of the most common and representative skin pathologies in clinical practice, including pressure ulcers, diabetic foot, water-soaked foot, fungal infections, chronic ulcers, eczema, atopic dermatitis, cellulitis, erysipelas, psoriasis, abscesses, and burns. This selection is based on clinical statistical data, as these twelve pathologies account for over 95% of clinical skin care cases.

[0058] In the application scenario of the virtual patient skin care teaching system with enhanced pathological features, each pathological feature is accurately described using four core dimensions. The texture feature dimension extracts the statistical characteristics of surface texture using a local binary pattern algorithm. Specifically, for the texture characteristics of pressure ulcers, the texture regularity coefficient of normal skin is 0.8-0.9, while the texture regularity coefficient of stage 4 pressure ulcers is reduced to 0.2-0.3. This quantitative description provides an accurate benchmark for subsequent parameter control.

[0059] Color feature dimensions are analyzed in the HSV color space, where the hue component H reflects the basic color tendency of the pathology, the saturation component S reflects the color purity, and the value component V reflects the brightness and darkness of the color. For example, in a virtual patient skin care teaching scenario, the HSV value range for normal skin is typically H∈[10°,30°], S∈[0.2,0.4], V∈[0.6,0.8], while the HSV value range for inflammatory pathology is H∈[0°,15°], S∈[0.6,0.9], V∈[0.4,0.7].

[0060] The pathology feature generation module 1 also includes a feature encoding unit 11, which is responsible for mapping the pathology features into a 512-dimensional latent vector space. In the specific implementation of the present invention for virtual patient skin care teaching with enhanced pathology features, the dimensionality of the latent vector is allocated according to the principles of medical importance and feature complexity. Texture features, as the most intuitive visual representation, are allocated the first 128 dimensions. This dimension selection is based on the theoretical basis of texture analysis and can fully express the changes in various texture patterns. Color features are allocated to dimensions 129 to 256, shape features are allocated to dimensions 257 to 384, and severity is allocated to dimensions 385 to 512.

[0061] Preferably, the feature encoding unit 11 uses an improved StyleGAN3 architecture to implement the mapping of parameters to latent vectors. In the virtual patient skin care teaching system with enhanced pathological features, the mapping process can be expressed as:

[0062] ,

[0063] in The generated latent vector is a five-hundred-twelve-dimensional real number vector, which is used to represent the complete encoding information of the pathological features; is the texture feature parameter vector, which is a 128-dimensional real number vector containing various statistical features of the skin surface texture; is the color feature parameter vector, which is a 128-dimensional real number vector that describes the distribution characteristics of skin color in the HSV space; is the shape feature parameter vector, which is a 128-dimensional real number vector representing the geometric shape characteristics of the pathological area; is the severity parameter vector, which is a one hundred and twenty-eight-dimensional real number vector that quantifies the development stage of the pathology and the degree of damage; is the encoder mapping function, which is a nonlinear transformation function based on the StyleGAN3 architecture.

[0064] The multi-scale generation unit 12 is data-connected to the feature encoding unit 11 to achieve the generation of pathological images at three resolution levels. During the implementation of the virtual patient skin care teaching system with enhanced pathological features of the present invention, the low-resolution level is set to 64×64 pixels, which is mainly used to capture the global features of the pathology, such as the overall color distribution and the general shape outline. This level provides a quick preview function for the teaching system. The medium-resolution level is set to 128×128 pixels, which is responsible for expressing local texture details and boundary features, and providing students with medium-precision observation details. The high-resolution level reaches 400×512 pixels, which can show the most delicate pathological structures and surface features, and provide support for the precise diagnosis training of professional medical staff.

[0065] like Figure 3 As shown, parameter control module 2 provides an intuitive slider control interface, enabling real-time visual editing of pathological features. In a preferred embodiment of the present invention's virtual patient skin care teaching system with enhanced pathological features, the slider interface utilizes a five-segment color mapping system, based on standard medical pathology grading practices. The green segment corresponds to slider positions 0.0-0.2, representing normal skin condition; the yellow segment corresponds to positions 0.2-0.4, representing stage I lesions; the orange segment corresponds to positions 0.4-0.6, representing stage II lesions; the red segment corresponds to positions 0.6-0.8, representing stage III lesions; and the dark red segment corresponds to positions 0.8-1.0, representing stage IV lesions.

[0066] The parameter linkage unit 21 establishes strict medical constraints to ensure that the generated pathological features conform to medical principles. In a virtual patient skin care teaching scenario, for example, the physiological constraint stipulates that there is a negative correlation between the degree of tissue necrosis and the blood circulation status. When the degree of tissue necrosis exceeds 0.7, the blood circulation status must be below 0.3. The temporal constraint ensures that the pathological development follows the natural process, and acute inflammation cannot jump directly to the chronic healing stage. The severity consistency constraint ensures that different manifestation parameters of the same pathology are consistent in severity. For example, when the pain intensity is set to 0.8, the tissue damage level should also be above 0.7.

[0067] The interpolation algorithm unit 22 implements intelligent parameter interpolation strategy selection. In the virtual patient skin care teaching system with enhanced pathological features, the linear interpolation method is used for continuously changing features such as color gradients:

[0068] ,

[0069] in: Is the result of linear interpolation, indicating the position of the slider The interpolated value at ; is the starting value, which indicates the value of the slider at its initial position; is the end value, which indicates the value of the slider at the end position; is the slider position parameter, where Corresponding to the leftmost end of the slider, The rightmost end of the slider corresponds to the operator ·, which represents a scalar multiplication operation. For parameters with threshold effects such as pain perception, nonlinear interpolation is used in the virtual patient skin care teaching system:

[0070] ,

[0071] in is the nonlinear interpolation result, which indicates the interpolation value after considering the threshold effect; It is the adjustment parameter of the starting value, usually set To simulate the nonlinear response of biological systems; It is the adjustment parameter of the termination value, usually set To reflect the characteristics of threshold effect; A nonlinear weight function representing the starting value; A nonlinear weight function representing the termination value.

[0072] When multiple sliders are adjusted simultaneously, the interpolation algorithm unit 22 implements multi-feature fusion through a priority allocation mechanism and dynamic weight adjustment. In the virtual patient skin care teaching scenario with enhanced pathological features, the multi-feature fusion weight allocation is calculated as:

[0073] ,

[0074] in: is the multi-feature fusion result vector; is the number of sliders that can be adjusted simultaneously; For the The weight coefficient of the feature satisfies ; For the feature vectors; For the The current position value of each slider; Indicates a weighted sum operation on all features.

[0075] like Figure 4As shown, the physical modeling module 3 simulates the physical effects of exudate reflection and stratum corneum exfoliation. In the virtual patient skin care teaching system with enhanced pathological features, the optical property calculation unit 31 calculates the refractive index of the exudate based on its composition and concentration. In a specific implementation of the present invention, inflammatory exudate, due to its high protein content, typically has a refractive index between 1.35 and 1.38. Infectious exudate, due to the presence of bacteria and pus, has a turbid appearance and a refractive index of approximately 1.40-1.42. Hemorrhagic exudate contains red blood cells and exhibits a distinctive reddish sheen, with a refractive index of approximately 1.38-1.40.

[0076] The Fresnel reflection calculation unit 32 processes the reflection phenomenon of light at the interface of different media. In the virtual patient skin care teaching system, the Fresnel reflection coefficient calculation formula implemented by this unit is:

[0077] ,

[0078] in: is the Fresnel reflection coefficient, which represents the ratio of the reflected light intensity to the incident light intensity, and its value range is [0,1]; is the refractive index of the incident medium, for air medium ; is the refractive index of the exudate medium, which is in the range of [1.35,1.42] according to the type of exudate; is the angle of incidence, which represents the angle between the incident light and the interface normal, in radians; is the refraction angle, which represents the angle between the refracted light and the interface normal, in radians, satisfying Snell's law ; is the cosine function; Indicates absolute value operation; superscript 2 indicates square operation; the first term is the reflection coefficient of s-polarized light; the second term is the reflection coefficient of p-polarized light; It means taking the average of the reflection coefficients of the two polarized lights.

[0079] The peeling probability calculation unit 33 establishes a stratum corneum peeling model based on stress analysis. In the virtual patient skin care teaching system with enhanced pathological characteristics, this unit calculates the peeling probability of each pixel:

[0080] ,

[0081] in: For location In time The peeling probability ranges from , the larger the value, the higher the possibility of peeling; is the horizontal coordinate of the pixel, and its value range is Corresponding image width; is the vertical coordinate of the pixel, and its value range is Corresponding image height; is the time parameter, in seconds, indicating the cumulative time from the initial state; is the exfoliation rate constant, in units of per second. This value is determined based on experimental data and reflects the average rate of stratum corneum exfoliation; For location The stress intensity coefficient ranges from , the larger the value, the greater the stress at that location; is a natural exponential function; the negative sign indicates that the parameter of the exponential function is negative, ensuring that the probability of spalling gradually increases and approaches 1 with time.

[0082] like Figure 5 As shown in Figure 4, the intelligent evaluation module 4 establishes a multi-dimensional evaluation system including basic indicators and judgment indicators. In the virtual patient skin care teaching system with enhanced pathological characteristics, the time series analysis unit 41 adopts a multi-frequency data acquisition strategy to ensure that the acquisition frequency of different types of indicators matches their change characteristics. Physiological parameters such as body temperature and heart rate are collected at a high frequency of once per minute, corresponding to an acquisition frequency of Morphological parameters such as swelling degree and color change are collected by medium frequency once an hour, and the corresponding collection frequency is Development parameters such as healing progress and tissue repair are collected at a low frequency once a day, corresponding to a collection frequency of .

[0083] The trend prediction unit 42 realizes the intelligent identification of short-term and long-term trends. In the virtual patient skin care teaching system, the short-term trend analysis is based on the data of the last 48 hours, using the moving average algorithm:

[0084] ,

[0085] in: For the moment The short-term trend value indicates the average level of observations in the recent period; is the short-term window length, which means considering the data of the latest 48 time points; For the The observed value at time is the time offset; Indicates moving forward from the current moment Sum all observations at each time point; is the normalization coefficient used to calculate the mean value.

[0086] Long-term trend forecasting uses an exponential smoothing algorithm:

[0087] ,

[0088] in: For the moment The long-term trend value reflects the long-term trend of change; is the smoothing coefficient, and its value range is (0,1). The smaller the value, the higher the degree of dependence on historical data. is the observation value at the current moment; is the long-term trend value of the previous moment; is the weight coefficient of historical trend.

[0089] The threshold judgment criteria of the five-level warning system are defined in the virtual patient skin care teaching system with enhanced pathological characteristics as follows:

[0090] ,

[0091] in: is the warning level determined by the evaluation score; score is the comprehensive evaluation score, and its value range is [0,1]. The conditional judgment uses the left-closed and right-open interval to ensure that each score value has a unique corresponding warning level.

[0092] Multimodal feedback module 5 achieves data fusion across three modalities: vision, touch, and behavior. In the virtual patient skin care teaching system with enhanced pathology features, the visual feature extraction unit extracts multi-level features from pathology images. Low-level visual features are extracted using a convolutional neural network, resulting in a 256-dimensional feature vector. Mid-level visual features are obtained using a shape analysis algorithm, resulting in a 128-dimensional feature vector. High-level semantic features are extracted using a pretrained classification network, resulting in a 512-dimensional feature vector.

[0093] The tactile signal processing unit analyzes the multi-dimensional signals from the tactile feedback device. In the virtual patient skin care teaching system, the pressure signal acquisition frequency is Hz, ensuring that rapid pressure changes can be captured; the acquisition frequency of the vibration signal is Hz, which meets the high frequency requirement of vibration perception; the acquisition frequency of temperature signal is Hz matches the time scale of temperature changes.

[0094] Multimodal data fusion uses a weighted fusion algorithm:

[0095] ,

[0096] in: is the fused feature vector, which is an 896-dimensional real vector; is the visual feature vector, which is a 256-dimensional real number vector containing the visual features extracted from the pathological image; is the tactile feature vector, which is a 128-dimensional real vector containing pressure, vibration and temperature information; is the behavior feature vector, which is a 512-dimensional real number vector containing user operation sequence and interaction mode information; is the weight coefficient of the visual modality; is the weight coefficient of the tactile modality; is the weight coefficient of the behavioral mode; the weight coefficient satisfies the normalization condition .

[0097] Adaptive Optimization Module 6 implements personalized adaptation based on user characteristics. In the virtual patient skin care teaching system enhanced with pathological features, the user modeling unit identifies the user's learning style, cognitive level, and emotional state. Learning style identification is based on the user's dwell time and interaction frequency on different types of learning materials, using a naive Bayesian classifier.

[0098] The strategy generation unit generates personalized feedback strategies based on user characteristics. In the virtual patient skin care teaching system, strategy generation uses a reinforcement learning algorithm, and the reward function is defined as:

[0099] ,

[0100] in In state Take action The reward value range is is the current system status, including user characteristics, learning progress and environment parameters; Actions taken for the system, including interface adjustments, content recommendations, and feedback strategies; is the accuracy score, with a value range of [0,1], reflecting the accuracy of the system output; The satisfaction score ranges from [0,1], reflecting the user's satisfaction with the system response; Efficiency score, with a value range of [0,1], reflecting the degree of improvement in learning efficiency; is the weight coefficient of accuracy; is the weight coefficient of satisfaction; is the efficiency weight coefficient; the weight coefficient satisfies the normalization condition .

[0101] The performance monitoring unit continuously monitors four key performance indicators. In the virtual patient skin care teaching system with enhanced pathological characteristics, the response time requirement is ms, accuracy requirement ,User satisfaction requirements (Based on a 10-point scale), resource utilization control To ensure system stability.

[0102] The present invention also includes extended components such as a skin light field database and a virtual reality rendering module. In the virtual patient skin care teaching system with enhanced pathological features, the skin light field database stores skin light field data covering different age groups, genders, and lesion locations. The database uses a hierarchical storage structure, categorizing skin light field data by age: children (0-12 years), youth (13-35 years), middle-aged (36-60 years), and elderly (over 60 years); by gender: male and female; and by body part: face, trunk, limbs, hands, and feet.

[0103] The virtual reality rendering module provides an immersive 3D visual experience through the Oculus Quest 2 helmet. In the virtual patient skin care teaching scene, the rendering engine uses the Unity 3D platform, supports real-time ray tracing and physical shaders, and the rendering frame rate is maintained at 90, ensuring a smooth user experience.

[0104] This invention utilizes a cloud computing architecture to provide powerful computing support. In a virtual patient skin care teaching system with enhanced pathological features, the cloud computing service module utilizes a GPU cluster to accelerate the StyleGAN3 and U-net computations. In actual deployment, a single user's image generation task is allocated 8GB of GPU memory, and batch processing tasks can handle 16 user requests simultaneously, with an average response time of less than 150 milliseconds.

[0105] The data synchronization module ensures data consistency between local and cloud servers. In the virtual patient skin care teaching system, the synchronization protocol uses an incremental synchronization mechanism, transmitting only the changed data, reducing network bandwidth usage. A data compression rate of 70% effectively reduces transmission latency.

[0106] The virtual patient skin care teaching method with enhanced pathological characteristics of the present invention includes five main steps, forming a complete teaching closed loop.

[0107] In step S1, the pathology feature library is constructed using raw image data acquired using a high-resolution microscope. In the virtual patient skin care teaching system with enhanced pathology features, the microscope's resolution is set to 2048 × 2048 pixels, and the magnification is 40x, ensuring clear observation of the skin's microstructure. Image processing includes denoising, enhancement, and normalization. Denoising uses a Gaussian filter with a standard deviation of σ = 1.5 pixels; enhancement uses histogram equalization; and normalization resizes the image to a uniform size of 400 × 512 pixels.

[0108] In step S2, the receipt and verification of user interaction parameters ensures system robustness. In the virtual patient skin care teaching scenario, parameter validation rules include numerical range checks, logical consistency checks, and medical plausibility checks. For example, the inflammation intensity parameter range is limited to [0.0, 1.0]. When the inflammation intensity exceeds 0.8, the system automatically sets the pain level to above 0.7.

[0109] In step S3, the computational complexity of the pathological image generation process is effectively controlled by GPU parallel computing. In the virtual patient skin care teaching system with enhanced pathological features, the forward propagation time of the StyleGAN3 encoder is approximately ms, the processing time of the U-net decoder is approximately ms, the physical effect addition time is about ms, total image generation time ms.

[0110] In step S4, the generation of multimodal feedback ensures the richness of user experience. In the virtual patient skin care teaching system, the rendering of visual feedback adopts delayed shading technology to support real-time shadow and reflection effects; the delay control of tactile feedback is Within 20ms, it meets the real-time requirements of tactile perception; the recording frequency of behavioral data is Hz, can accurately capture the user's operation details.

[0111] In step S5, intelligent evaluation and optimization form the adaptive ability of the system. In the virtual patient skin care teaching system with enhanced pathological characteristics, the evaluation algorithm is executed every 5 minutes to ensure that changes in the user's learning status can be discovered in time; the optimization algorithm adopts an online learning mechanism, and the learning rate is set to , ensuring that the system can quickly adapt to changes in user characteristics.

[0112] Through the above detailed description of the embodiments, those skilled in the art will be able to understand and implement the technical solutions of the present invention, thereby achieving significant technical effects and application value. The present invention not only solves the technical difficulties in traditional medical teaching, but also provides an important technical foundation for the future development of intelligent medical education.

[0113] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A virtual patient skin care teaching system with enhanced pathological features, characterized by: include: A pathology feature generation module is used to construct a library of twelve types of skin pathology features, including pressure sores, diabetic foot, water foot, fungal infection, chronic ulcers, eczema, atopic dermatitis, cellulitis, erysipelas, psoriasis, abscesses, and burns. Each pathology feature is vectorized and encoded using four dimensions: texture, color, shape, and severity. An encoder based on the StyleGAN3 architecture maps the pathology parameters into a 512-dimensional latent feature vector, which is then converted into a 400 by 512-pixel pathology skin image using a U-net architecture decoder. a parameter control module, data-connected to the pathology feature generation module, for providing a slider control interface to enable real-time visual editing of pathology features, wherein the slider adopts a five-segment color mapping system, with the green segment corresponding to normal skin condition, the yellow segment corresponding to stage I lesions, the orange segment corresponding to stage II lesions, the red segment corresponding to stage III lesions, and the dark red segment corresponding to stage IV lesions, and adjusting the values of corresponding dimensions in the latent feature vector in real time based on user slider operation; a physical modeling module, data-connected to the pathological feature generation module and the parameter control module, for constructing an exudate reflection physical model and a stratum corneum exfoliation dynamics model based on the pathological skin image, wherein the exudate reflection physical model simulates the optical reflection effect of the liquid surface by creating a binary mask and combining alpha blending technology, and the stratum corneum exfoliation dynamics model calculates the exfoliation probability at different locations through a probability distribution function and generates a progressive exfoliation animation; An intelligent assessment module, data-connected to the pathological feature generation module, parameter control module, and physical modeling module, is used to establish a multi-dimensional assessment system comprising basic indicators and judgment indicators. The basic indicators include four physiological parameters: body temperature, heart rate, blood pressure, and respiration; and four morphological parameters: color, texture, size, and depth. The judgment indicators include site risk assessment, severity quantification, and healing prediction. A comprehensive assessment score is generated by fusing various indicators through a weight matrix. A multimodal feedback module, data-connected to the intelligent evaluation module, is used to integrate visual feedback, tactile feedback, and user behavior feedback. Visual feedback is provided by pathological images generated by real-time rendering. Tactile feedback uses virtual touch sensors to detect user operations and generate corresponding vibration effects to simulate skin texture and roughness. User behavior feedback obtains user satisfaction and interest data by analyzing operation sequences, dwell time, and parameter adjustment frequency. An adaptive optimization module is data-connected to the multimodal feedback module and is used to dynamically adjust system parameters through a reinforcement learning algorithm based on the user satisfaction and interest data and the comprehensive evaluation score, and to provide personalized interface configuration and function recommendations based on the user's professional background, experience level, and usage scenarios.

2. The virtual patient skin care teaching system with enhanced pathological characteristics according to claim 1 is characterized in that: The pathological feature generation module also includes: a feature encoding unit, configured to map each dimension of the twelve types of skin pathology features to a specific segment of the 512-dimensional latent feature vector, wherein the texture feature dimension is mapped to the first to the 128th dimensions, the color feature dimension is mapped to the 129th to the 256th dimensions, the shape feature dimension is mapped to the 257th to the 384th dimensions, and the severity dimension is mapped to the 385th to the 512th dimensions; a multi-scale generation unit, data-connected to the feature encoding unit, for generating a pathological image at three resolution levels based on the potential feature vector, wherein a low resolution of sixty-four by sixty-four pixels is used for global shape and color distribution, a medium resolution of one hundred and twenty-eight by one hundred and twenty-eight pixels is used for local texture and boundary features, and a high resolution of four hundred by five hundred and twelve pixels is used for subtle pathological structures and surface features.

3. The virtual patient skin care teaching system with enhanced pathological features according to claim 1 is characterized in that: The parameter control module also includes: The parameter linkage unit is used to establish medical constraints between pathological parameters, including physiological constraints to prevent parameter combinations that are medically impossible to occur simultaneously, temporal constraints to ensure the temporal logic and rationality of pathological development, and severity consistency constraints to ensure that different manifestation parameters of the same pathology are consistent in severity. The interpolation algorithm unit is data-connected to the parameter linkage unit, and is used to automatically select linear interpolation, nonlinear interpolation or segmented interpolation strategies according to the pathological feature type and parameter properties, and to achieve multi-feature fusion through a priority allocation mechanism and dynamic weight adjustment when multiple sliders are adjusted simultaneously.

4. The virtual patient skin care teaching system with enhanced pathological features according to claim 1, characterized in that: The physical modeling module also includes: an optical property calculation unit, configured to calculate the refractive index based on the composition and concentration of the exudate, wherein the refractive index of inflammatory exudate is set between 1.35 and 1.38, infectious exudate has a turbid appearance, and bloody exudate has a red sheen, and to establish a thickness distribution model based on the microscopic morphology of the skin; A Fresnel reflection calculation unit, connected to the optical characteristic calculation unit, is used to process the differences in reflection characteristics under different incident angles, calculate the reflection coefficients of s-polarized light and p-polarized light respectively, and consider the multi-interface reflection effects of the air-exudate interface and the exudate-skin interface; The peeling probability calculation unit is used to analyze the stress distribution on the skin surface and establish a probability-based peeling model, in which local stress analysis determines the peeling possibility of each pixel point, the material fatigue model considers the impact of historical stress on the current peeling probability, and random factors simulate the uncertainty of the peeling process.

5. The virtual patient skin care teaching system with enhanced pathological features according to claim 1, characterized in that: The intelligent evaluation module further includes: A time series analysis unit is used to adopt a multi-frequency data acquisition strategy, in which physiological parameters are collected at a high frequency of once per minute, morphological parameters are collected at a medium frequency of once per hour, and developmental parameters are collected at a low frequency of once per day. Data quality control is achieved through outlier detection, missing value processing, and noise filtering; The trend prediction unit is connected to the data of the time series analysis unit and is used to identify short-term trends of 24 to 48 hours and long-term trends of several weeks to several months, detect trend turning points, and establish a five-level early warning system, where green indicates normal status, yellow indicates slight abnormality, orange indicates moderate abnormality, red indicates severe abnormality, and purple indicates extremely severe abnormality.

6. The virtual patient skin care teaching system with enhanced pathological features according to claim 1, characterized in that: The multimodal feedback module further includes: a visual feature extraction unit, configured to extract low-level visual features, mid-level visual features, and high-level semantic features from the pathological skin image, wherein the low-level visual features include color histogram, texture features, edge features, and gradient features; the mid-level visual features include shape features, symmetry features, connectivity features, and distribution features; and the high-level semantic features include pathology type classification, severity assessment, development stage judgment, and treatment effect evaluation; The tactile signal processing unit is used to analyze the pressure signals, vibration signals and temperature signals generated by the tactile feedback device. The pressure signal analysis includes the pressure magnitude, pressure distribution, pressure duration and pressure change rate. The vibration signal analysis includes the vibration frequency, vibration amplitude, vibration duration and vibration pattern. The temperature signal analysis includes the surface temperature, temperature change rate, temperature distribution pattern and temperature recovery time.

7. The virtual patient skin care teaching system with enhanced pathological features according to claim 1, characterized in that: The adaptive optimization module also includes: The user modeling unit is used to identify the user's learning style, cognitive level, and emotional state. Learning styles include visual, auditory, kinesthetic, and reading; cognitive levels are categorized as beginner, advanced, and expert; and emotional states include positive, negative, confused, and satisfied. A strategy generation unit, connected to the user modeling unit, is used to generate personalized feedback strategies based on user characteristics, including adjusting feedback methods for different learning styles, adjusting feedback complexity based on cognitive level, and adjusting the emotional color of feedback based on emotional state; The performance monitoring unit is used to continuously monitor four key performance indicators: response time, accuracy, user satisfaction, and resource utilization. It automatically identifies computing bottlenecks, storage bottlenecks, network bottlenecks, and user experience bottlenecks, and uses a predictive maintenance mechanism to detect and prevent potential problems in advance.

8. The virtual patient skin care teaching system with enhanced pathological features according to claim 1, characterized in that: Also includes: A skin light field database, connected to the pathological feature generation module, is used to store skin light field data of different age groups, genders, and lesion locations, including skin attributes of the face, trunk, limbs, hands, and feet, and corresponding light field parameter settings; The virtual reality rendering module is connected to the skin light field database and the physical modeling module data, and is used to render the three-dimensional virtual scene in real time through the Oculus Quest2 helmet and HMD head-mounted display, and generate the final visual output in combination with environmental parameters such as lighting conditions and observation angle.

9. The virtual patient skin care teaching system with enhanced pathological features according to claim 1, characterized in that: The system adopts a cloud computing architecture, including: A cloud computing service module is used to utilize cloud-based GPU computing resources to accelerate the computational process of the StyleGAN3 architecture encoder and the U-net architecture decoder, and dynamically allocate computing resources based on the number of users through elastic resource scheduling; The data synchronization module is connected to the cloud computing service module data, and is used to synchronize the pathological characteristic parameters, user operation records and evaluation result data between the local client and the cloud server, and ensure the consistency and integrity of the data through the version control mechanism.

10. A virtual patient skin care teaching method based on the pathological characteristics enhanced system of any one of claims 1 to 9, characterized in that: The following steps are involved: S1: Pathological feature library construction step, using a high-resolution microscope to collect raw image data of the twelve types of skin pathological features, and using image processing technology to extract feature parameters in four dimensions: texture feature, color feature, shape feature, and severity, to establish a standardized feature library containing the four development stages of each type of pathological feature; S2: a user interaction parameter setting step, receiving the pathology type selection and severity adjustment parameters input by the user through the slider control interface, verifying the rationality of the parameters according to the constraint rules of the parameter linkage unit, and calculating the corresponding potential feature vector value through the interpolation algorithm unit; S3: Pathological image generation step, inputting the potential feature vector into the encoder of the StyleGAN3 architecture for feature mapping, generating a basic pathological skin image through the decoder of the U-net architecture, and adding exudate reflection effect and stratum corneum peeling effect in combination with the physical modeling module; S4: a multimodal feedback generating step, generating three-dimensional visual feedback through the virtual reality rendering module based on the pathological skin image, generating a corresponding tactile feedback signal through the virtual touch sensor, and recording the user's operation behavior data; S5: Intelligent evaluation and optimization step, using the intelligent evaluation module to perform multi-dimensional evaluation on the generated pathological images and generate a comprehensive evaluation score, using the adaptive optimization module to analyze user feedback data and adjust system parameters, and provide personalized learning suggestions and system optimization solutions based on the evaluation results and user characteristics.

Images