A Computer Vision-Based Assessment System and Method for Facial Acne Atrophic Scars

Abstract

This invention relates to the field of facial acne repair technology, and more particularly to a computer vision-based system and method for assessing facial acne atrophic scars. The system includes: analyzing the intensity of facial pollution behaviors caused by the patient's daily activities based on patient daily behavior data and scar location distribution data; predicting the patient's risk of stress response in the scar area under future environmental conditions by combining environmental status data with epidermal morphology data of the patient's facial acne atrophic scar area; assessing the risk of scar recovery deterioration during the patient's scar repair process based on the results of the facial pollution behavior intensity analysis and the predicted risk of stress response in the scar area; and providing early warning for medical follow-up examinations based on the results of the scar recovery deterioration risk assessment. This system can accurately assess the state of acne atrophic scars, dynamically monitor the risk of recovery deterioration, and promptly trigger follow-up examination warnings, effectively improving the scar repair effect and the accuracy of medical intervention.

Description

Technical Field

[0001] This invention relates to the field of facial acne repair technology, and in particular to a computer vision-based system and method for assessing facial acne pitted scars. Background Technology

[0002] Facial atrophic acne scars, a common secondary skin problem following acne inflammation healing, not only damage the integrity of facial appearance but also significantly impact patients' mental health. Meanwhile, the incidence of facial acne remains consistently high, with many patients developing ice pick, rolling, or boxcar-shaped atrophic scars due to factors such as the depth of inflammation, improper intervention during the healing process (e.g., picking, squeezing, or inadequate cleaning). Once these scars form, the treatment may be ineffective due to a mismatch between the intensity of repair and the severity of the scar, wasting medical resources. In severe cases, delayed early intervention can lead to further thinning of the scar epidermis, continued collagen loss, and even secondary inflammation caused by repeated contamination, creating a vicious cycle. Furthermore, the effectiveness of scar repair requires real-time monitoring of the scar's epidermal micromorphology, location distribution, and the risks of contamination from the patient's daily activities, as well as stress responses from environmental factors. These factors collectively determine the scar's recovery trend. By introducing computer vision to extract the three-dimensional coordinates and epidermal morphological features of the scar area, the repair effect of facial atrophic acne scars can be improved to some extent.

[0003] However, current technologies for scar repair only count the total number of hand contacts using motion capture, without analyzing the location and distribution characteristics of the scar. If hand contact is concentrated in the cheek scar area, the risk of contamination is far higher than contact with normal skin on the forehead. However, current technologies only count the total number of contacts, making it impossible to accurately calculate the intensity of facial contamination and thus misjudge the risk of scar deterioration due to contamination. Furthermore, current technologies lack the ability to collaboratively analyze environmental data and scar epidermal morphology data. Ultraviolet radiation damages the stratum corneum structure of the scar epidermis and inhibits fibroblast activity. Air pollutants can penetrate the damaged barrier into the dermis, exacerbating the inflammatory response. However, current technologies either assess environmental risks in isolation or only analyze epidermal morphology, without analyzing the interaction between environmental parameters and epidermal morphological parameters. This results in low accuracy in predicting future scar deterioration trends and an inability to provide patients with forward-looking environmental protection recommendations.

[0004] To address these issues, this application presents a computer vision-based system and method for assessing facial acne pitted scars. Summary of the Invention

[0005] The purpose of this invention is to provide a computer vision-based assessment system and method for facial acne atrophic scars. Supported by computer vision technology, it acquires data on the epidermal morphology and location distribution of facial acne atrophic scars, as well as data on patients' daily behavior and environmental conditions. It constructs pollution behavior intensity analysis, scar stress response risk prediction, and dynamic assessment model of recovery and deterioration risk. Based on the deterioration risk assessment results, it implements medical follow-up warnings for patients, thereby achieving accurate dynamic assessment and intervention prompts for scar repair.

[0006] This invention is implemented as follows:

[0007] In a first aspect, this invention provides a computer vision-based assessment system for facial acne atrophic scars, including a data acquisition module, a facial pollution behavior intensity analysis module, a scar area stress response risk prediction module, a scar recovery and deterioration risk dynamic assessment module, and a medical follow-up early warning module. The data acquisition module acquires epidermal morphology data and scar location distribution data of the patient's facial acne atrophic scar area, as well as patient daily behavior data and environmental status data. The facial pollution behavior intensity analysis module, based on the patient's daily behavior data and scar location distribution data, constructs a facial pollution behavior intensity analysis model using computer vision technology to assess the facial pollution caused by the patient's daily behavior. The system analyzes the intensity of pollution behavior; the scar area stress response risk prediction module combines environmental status data with epidermal morphology data of the patient's facial acne atrophic scar area to construct a scar area stress response risk prediction model, predicting the patient's scar area stress response risk in future environments; the scar recovery deterioration risk dynamic assessment module, based on the facial pollution behavior intensity analysis results and scar area stress response risk prediction results, constructs a scar recovery deterioration risk dynamic assessment model, dynamically assessing the patient's scar recovery deterioration risk during the scar repair process; the medical follow-up early warning module, based on the dynamic assessment results of scar recovery deterioration risk, provides early warnings for patients requiring medical follow-up.

[0008] In one implementation of the present invention, a facial contamination behavior intensity analysis model is constructed using computer vision technology to analyze the intensity of facial contamination behavior caused by the patient's daily behavior, including the following specific steps:

[0009] S21. Extract the total number of times the hand touches the face and the duration of each touch from the patient's daily behavior data during the current monitoring period; use the ratio of the total number of times the hand touches the face to the number of monitoring days as the daily average contact frequency, and use the average duration of each touch as the average single contact duration; multiply the daily average contact frequency and the average single contact duration to obtain the behavioral contact intensity.

[0010] S22. Extract the total area of ​​the scar region from the scar location distribution data, and extract the overlapping area between the hand contact area and the scar region within the current monitoring period from the daily behavior data; take the ratio of the overlapping area to the total hand contact area within the monitoring period as the scar region contact ratio; take the ratio of the overlapping area to the total scar region area as the scar region coverage ratio; and calculate the arithmetic mean of the scar region contact ratio and the scar region coverage ratio to obtain the location association strength.

[0011] S23. Perform a geometric mean calculation on the behavioral contact intensity and the location association intensity, and use the result of the geometric mean calculation as the facial contamination behavioral intensity caused by the patient's daily behavior.

[0012] In one implementation of the present invention, a scar area stress response risk prediction model is constructed to predict the patient's scar area stress response risk in future environments, including the following specific steps:

[0013] S31. Extract the ultraviolet index and air pollutant index for a certain historical monitoring period from the environmental status data; calculate the arithmetic average of the ultraviolet index and the air pollutant index to obtain the environmental risk coefficient for a certain historical monitoring period.

[0014] S32. Extract the epidermal thickness and skin barrier function index of the patient's scar area from the epidermal morphology data within a certain historical monitoring period; simultaneously obtain the average epidermal thickness and average skin barrier function index of the patient's healthy facial skin area within a certain historical monitoring period; divide the epidermal thickness of the patient's scar area by the average epidermal thickness of the patient's healthy facial skin area to obtain the relative thinness of the epidermis; divide the skin barrier function index of the patient's scar area by the average skin barrier function index of the patient's healthy facial skin area to obtain the relative damage to the barrier; perform a geometric mean operation on the relative thinness of the epidermis and the relative damage to the barrier, and use the reciprocal of the geometric mean operation result as the epidermal damage risk base; extract the ratio of the total area of ​​the scar area to the total area of ​​the face from the scar location distribution data as the scar area proportion; multiply the epidermal damage risk base by the scar area proportion, and use the multiplication result as the stress response risk of the scar area within a certain historical monitoring period.

[0015] In one implementation of the present invention, a scar area stress response risk prediction model is constructed to predict the patient's scar area stress response risk in future environments, and the following specific steps are also included:

[0016] S33. Based on the calculation process in step S31, obtain the environmental risk coefficient within multiple historical monitoring cycles; based on the calculation process in step S32, obtain the stress response risk of the scar area within multiple historical monitoring cycles; use the environmental risk coefficient within a certain historical monitoring cycle and the stress response risk of the scar area within the previous historical monitoring cycle corresponding to that historical monitoring cycle as the input dataset for that historical monitoring cycle; use the input datasets of multiple historical monitoring cycles and the corresponding stress response risks of the scar area in multiple historical monitoring cycles as the paired sample dataset; divide the paired sample dataset into a paired sample training set and a paired sample test set in a 7:3 ratio.

[0017] S34. Construct a gradient boosting regression tree model. Use the input dataset of all historical monitoring cycles in the paired sample training set as the input features of the gradient boosting regression tree model. Use the scar area stress response risk of all historical monitoring cycles in the paired sample training set as the prediction target of the gradient boosting regression tree model. Train the gradient boosting regression tree model to obtain the initial scar area stress response risk prediction model.

[0018] S35. The initial scar area stress response risk prediction model is tested using a paired sample test set to obtain the scar area stress response risk prediction model.

[0019] S36. Obtain the scar area stress response risk within the current monitoring period, and simultaneously extract the predicted ultraviolet index and air pollutant index for the next monitoring period from the environmental status data to calculate the environmental risk coefficient for the next monitoring period; input the scar area stress response risk within the current monitoring period and the environmental risk coefficient for the next monitoring period into the scar area stress response risk prediction model for prediction, and use the predicted value output by the scar area stress response risk prediction model as the patient's scar area stress response risk in the future environment.

[0020] In one implementation of the present invention, a dynamic evaluation model for scar recovery effect is constructed to assess the risk of scar deterioration during the patient's scar repair process, including the following specific steps:

[0021] S41. Obtain the results of the intensity analysis of facial pollution behavior caused by the patient's daily behavior and the results of the prediction of the patient's risk of stress response in the scar area in the future environment;

[0022] S42. The comprehensive risk index is obtained by geometrically averaging the results of facial pollution behavior intensity analysis and the results of scar area stress response risk prediction.

[0023] S43. Extract the average epidermal repair rate and average collagen density of the scar area within the current monitoring period from the epidermal morphology data. Simultaneously, obtain the average epidermal repair rate and average collagen density of the healthy facial skin area within the current monitoring period. Divide the average epidermal repair rate of the scar area within the current monitoring period by the average epidermal repair rate of the healthy facial skin area within the current monitoring period to obtain the repair rate ratio. Divide the average collagen density of the scar area within the current monitoring period by the average collagen density of the healthy facial skin area within the current monitoring period to obtain the collagen density ratio. Perform a geometric mean calculation on the repair rate ratio and the collagen density ratio, and use the result of the geometric mean calculation as the patient's scar self-repair ability status.

[0024] S44. Divide the comprehensive risk index by the patient's scar self-repair ability and use the result of the division as the risk of scar deterioration during the patient's scar repair process.

[0025] In one implementation of the present invention, based on the risk assessment results of scar recovery deterioration, a medical follow-up warning is issued to the patient, including the following specific steps:

[0026] S51. Set a medical follow-up examination early warning threshold and obtain the results of the scar recovery and deterioration risk assessment during the patient's scar repair process;

[0027] S52. Compare the risk assessment results of scar recovery deterioration during the patient's scar repair process with the medical follow-up warning threshold; if the risk assessment results of scar recovery deterioration during the patient's scar repair process are greater than the medical follow-up warning threshold, generate a medical follow-up warning signal for the patient; if the risk assessment results of scar recovery deterioration during the patient's scar repair process are less than or equal to the medical follow-up warning threshold, continue to monitor the patient's scar repair process.

[0028] Secondly, the present invention provides a computer vision-based method for assessing facial acne atrophic scars, comprising the following specific steps:

[0029] S1. Obtain epidermal morphology data and scar location distribution data of the patient's facial acne pitted scar area, and at the same time obtain the patient's daily behavior data and environmental status data;

[0030] S2. Based on patients' daily behavior data and scar location distribution data, a facial contamination behavior intensity analysis model is constructed using computer vision technology to analyze the intensity of facial contamination behavior caused by patients' daily behavior.

[0031] S3. By combining environmental data with epidermal morphology data of the patient's facial acne pitted scar area, a scar area stress response risk prediction model is constructed to predict the patient's scar area stress response risk in future environments.

[0032] S4. Based on the analysis results of facial pollution behavior intensity and the prediction results of stress response risk in scar area, a dynamic assessment model for the risk of scar recovery deterioration is constructed to assess the risk of scar recovery deterioration during the patient's scar repair process;

[0033] S5. Based on the risk assessment results of scar recovery and deterioration, conduct medical follow-up examinations and provide early warnings for patients.

[0034] Thirdly, the present invention provides an electronic device, comprising: a processor and a memory, wherein the memory stores a computer program that can be called by the processor, and the processor executes a computer vision-based method for assessing facial acne pitted scars by calling the computer program stored in the memory.

[0035] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0036] This invention analyzes the intensity of facial pollution caused by patients' daily behaviors based on patient daily behavior data and scar location distribution data; combines environmental status data with epidermal morphology data of the patient's facial acne atrophic scar area to predict the patient's risk of scar area stress response in future environments; based on the results of facial pollution behavior intensity analysis and scar area stress response risk prediction, assesses the risk of scar recovery deterioration during the patient's scar repair process; and provides early warning for medical follow-up examinations based on the scar recovery deterioration risk assessment results. This invention can accurately assess the state of acne atrophic scars, dynamically monitor the risk of recovery deterioration, and promptly trigger follow-up examination warnings, effectively improving scar repair outcomes and the accuracy of medical interventions. Attached Figure Description

[0037] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0038] Figure 1 This is a schematic diagram of the overall process of the computer vision-based facial acne pitted scar assessment method of the present invention;

[0039] Figure 2 This is a schematic diagram of the structure of the computer vision-based facial acne pitted scar assessment system of the present invention;

[0040] Figure 3 This is a flowchart illustrating step S2 of the computer vision-based facial acne pitted scar assessment method of the present invention. Detailed Implementation

[0041] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0042] Example 1

[0043] like Figure 1 As shown, this embodiment provides a computer vision-based method for assessing facial acne atrophic scars, including the following specific steps:

[0044] S1. Obtain epidermal morphology data and scar location distribution data of the patient's facial acne pitted scar area, and at the same time obtain the patient's daily behavior data and environmental status data;

[0045] S2. Based on patients' daily behavior data and scar location distribution data, a facial contamination behavior intensity analysis model is constructed using computer vision technology to analyze the intensity of facial contamination behavior caused by patients' daily behavior.

[0046] S3. By combining environmental data with epidermal morphology data of the patient's facial acne pitted scar area, a scar area stress response risk prediction model is constructed to predict the patient's scar area stress response risk in future environments.

[0047] S4. Based on the analysis results of facial pollution behavior intensity and the prediction results of stress response risk in scar area, a dynamic assessment model for the risk of scar recovery deterioration is constructed to assess the risk of scar recovery deterioration during the patient's scar repair process;

[0048] S5. Based on the risk assessment results of scar recovery and deterioration, conduct medical follow-up examinations and provide early warnings for patients.

[0049] In this embodiment, as Figure 3 As shown, step S2 uses computer vision technology to construct a facial contamination behavior intensity analysis model to analyze the intensity of facial contamination behavior caused by the patient's daily behavior, including the following specific steps:

[0050] S21. Extract the total number of hand-to-face contacts and the duration of each contact within the current monitoring period from the patient's daily behavior data; use the ratio of the total number of hand-to-face contacts to the number of monitoring days as the daily average contact frequency, and the average duration of each contact as the average single contact duration; multiply the daily average contact frequency and the average single contact duration to obtain the behavioral contact intensity; this embodiment systematically extracts the total number of hand-to-face contacts and the duration of each contact from the patient's daily behavior data; to ensure the objectivity and clinical reference value of the parameters, in this embodiment, the acquisition of the total number of hand-to-face contacts requires the deployment of multi-view motion capture. The system deploys 3-4 high-frame-rate cameras in key scenarios involving daily activities during the patient's scar repair process. Each camera samples at 60 frames per second. Combined with an AI skeletal keypoint recognition algorithm, it marks 12 key anatomical points, including the patient's wrist, index fingertip, middle fingertip, nasal ala, mandibular border, and brow ridge. By calculating the Euclidean distance between hand keypoints (e.g., fingertips in contact with the face) and facial keypoints (in this embodiment, the nasal ala is the default facial keypoint), a valid contact is defined as a distance less than 5mm that lasts for more than 0.3 seconds. This eliminates false positives in non-contact scenarios such as drinking water or tidying hair. In this embodiment, the monitoring period is set to 7-14 days to cover behavioral differences between weekdays and weekends. For example, some patients frequently touch their faces during weekdays due to stress, while their behavior is more relaxed at home on weekends. The 7-14 day data comprehensively reflects this behavioral cycle. In this embodiment, the duration of each contact is recorded using the system's timestamp function. The time difference from the start frame of contact to the end frame is the duration of a single contact. The sum of all single contact durations is divided by the total number of contacts to obtain the average duration of a single contact. The total number of contacts is divided by the number of monitoring days to obtain the average daily contact frequency. For example, acne scar patients may have an average daily contact frequency of 15-20 times due to itching from the scars, which is much higher than the 3-5 times of healthy individuals. The higher the frequency of hand contact and the longer the duration of a single contact, the longer the contaminants (such as Staphylococcus aureus, bacteria on the mobile phone surface, and environmental dust) remain on the face. The depressed structure of atrophic scars easily accumulates contaminants, causing local inflammation and damaging the epidermal barrier. Multiplying the average daily contact frequency by the average duration of a single contact quantifies the behavioral contact intensity when the patient's fingers come into contact with facial acne scars. In this embodiment, computer vision is used to analyze the contact behavior between the patient's fingers and facial acne scars. This avoids the subjective bias of patients concealing their face-picking behavior in traditional questionnaires and objectively restores the behavioral scenario. The quantified behavioral contact intensity provides a basis for subsequent contamination risk analysis. Doctors can use this to identify high-risk behaviors and provide targeted intervention suggestions to reduce scar contamination from the source and help with repair.

[0051] S22. Extract the total area of ​​the scar region from the scar location distribution data, and extract the overlapping area between the hand contact area and the scar region within the current monitoring period from daily behavior data; use the ratio of the overlapping area to the total hand contact area within the monitoring period as the scar region contact ratio; use the ratio of the overlapping area to the total scar region area as the scar region coverage ratio; perform an arithmetic mean calculation on the scar region contact ratio and the scar region coverage ratio to obtain the location association strength; this embodiment quantifies the targeted impact of hand contact on the scar region by accurately obtaining the spatial association parameters between the scar region and the hand contact area. In this embodiment, the total area of ​​the scar region is obtained using a structured light 3D scanner. Before scanning, the patient is fixed in a seated position with a headrest. The scanning range covers the entire face area from the hairline to the jawline, and from the left and right sides to the tragus. The scanning resolution is set to 0.1mm, generating a 3D point cloud model of the patient's entire face. Subsequently, the U-Net++ semantic segmentation algorithm was used to segment the scar region of the patient's full-face 3D point cloud model. In this embodiment, the training set of the U-Net++ semantic segmentation algorithm was based on 500 clinically labeled 3D images of acne atrophic scars, which can automatically identify and segment the boundary of each scar. The total area of ​​the scar region is obtained by summing the point cloud area using a point cloud area calculation algorithm. For example, the total scar area of ​​a patient with moderate acne scars may reach 8-12 square centimeters, and the distribution of ice pick scars and rolling scars is different. The segmentation algorithm in this embodiment can accurately distinguish the above-mentioned scar regions and different types of scars. In this embodiment, the process of obtaining the overlapping area between the hand contact area and the scar area is as follows: The 3D coordinate range of the daily hand contact with the face recorded by the motion capture system is superimposed on the 3D point cloud model of the face. The point cloud area of ​​the overlapping part is calculated by spatial intersection operation, which is the overlapping area; the total hand contact area is the sum of the areas of all hand contact areas on the patient's face. The scar area contact ratio is calculated by dividing the overlapping area by the total hand contact area. In this embodiment, the scar area contact ratio reflects the probability that the contact point falls on the scar. The scar area coverage ratio is calculated by dividing the overlapping area by the total scar area, reflecting the degree to which the scar is covered by contact. In this embodiment, the two are arithmetically averaged to obtain the location association strength. For example, if a patient's hand contact is mostly concentrated in the cheek scar area, their scar area contact ratio may reach 60%, and the location association strength is significantly higher than that of a patient whose contact is concentrated in normal skin. This embodiment can accurately locate the targeted impact of hand contact on scars. For example, if two patients have the same daily contact frequency, but one's contact is concentrated in the scar area, their contamination risk is much higher than that of a patient who contacts normal skin. The quantification of location association strength allows doctors to clearly identify the patient's high-risk contact areas and provides spatial weights for the comprehensive calculation of subsequent contamination behavior intensity, making the assessment more in line with the clinical needs of scar repair.

[0052] S23. Perform a geometric mean calculation on the behavioral contact intensity and the location association intensity, and use the result of the geometric mean calculation as the facial contamination behavioral intensity caused by the patient's daily behavior. This embodiment uses a geometric mean calculation to combine behavioral contact intensity and location association intensity, balancing the weights of the two to avoid assessment bias caused by one indicator being too high or too low; for example, if a patient has high behavioral contact intensity but low location association intensity because the contact is mostly on normal skin, the patient's facial contamination behavioral intensity will not be excessively high after geometric mean calculation; conversely, for patients with moderate contact intensity but high location association intensity due to concentrated contact in scar areas, the high facial contamination behavioral intensity can be reflected by geometric mean calculation.

[0053] In this embodiment, step S3 involves constructing a scar area stress response risk prediction model to predict the patient's scar area stress response risk in future environments. This includes the following specific steps:

[0054] S31. Extract the UV index and air pollutant index for a specific historical monitoring period from environmental status data; calculate the arithmetic mean of the UV index and air pollutant index to obtain the environmental risk coefficient for that historical monitoring period. This embodiment accurately obtains the daily UV index and air pollutant index from environmental status data for a specific monitoring period, and takes the average of all daily UV and air pollutant indices as the UV index and air pollutant index for that historical monitoring period, calculating the environmental risk coefficient for that period. This provides an objective quantitative basis for predicting the risk of stress response in the scar area. In this embodiment, UV and air pollutants are the two core environmental factors leading to scar stress response, and their synergistic effect exacerbates scar damage. This embodiment objectively quantifies the intensity of environmental stimuli, avoiding the ambiguity of patients' subjective descriptions of bad weather; and provides reliable input for predicting the risk of stress response by quantifying the environmental risk coefficient.

[0055] S32. Extract the epidermal thickness and skin barrier function index of the patient's scar area from the epidermal morphology data within a certain historical monitoring period; simultaneously obtain the average epidermal thickness and average skin barrier function index of the patient's healthy facial skin area within a certain historical monitoring period; divide the epidermal thickness of the patient's scar area by the average epidermal thickness of the patient's healthy facial skin area to obtain the relative thinness of the epidermis; divide the skin barrier function index of the patient's scar area by the average skin barrier function index of the patient's healthy facial skin area to obtain the relative damage to the barrier; perform a geometric mean operation on the relative thinness of the epidermis and the relative damage to the barrier, and use the reciprocal of the geometric mean operation result as the epidermal damage risk base; extract the ratio of the total area of ​​the scar area to the total area of ​​the face from the scar location distribution data as the scar area proportion; multiply the epidermal damage risk base by the scar area proportion, and use the multiplication result as the stress response risk of the scar area within a certain historical monitoring period.

[0056] For example, in this embodiment, the epidermal thickness value is obtained by using a confocal laser scanning microscope to acquire three-dimensional tomographic images of the skin from the epidermis to the superficial dermis in a non-invasive state, avoiding secondary damage to scar tissue caused by traditional biopsies. In selecting detection points in the scar area, different types of atrophic acne scars are covered, with 3-5 detection points selected for each type. For example, for ice pick scars, two points are selected at the bottom and two at the edge of the depression; for rolling scars, three points are selected in the gently sloping area of ​​the depression. During scanning, the probe is gently placed against the skin surface, and the scanning depth is set from the surface of the stratum corneum to the basal layer. The device's built-in image analysis software automatically identifies the boundaries between the stratum corneum and the granular layer, the granular layer and the spinous layer, and the spinous layer and the basal layer. The vertical distance from the surface of the stratum corneum to the basal layer is measured, which is the epidermal thickness value of a single detection point. The thickness values ​​of all detection points in the scar area are taken as an arithmetic mean to obtain the epidermal thickness value of the patient's scar area within this historical monitoring period. To obtain the average epidermal thickness of healthy facial skin, a control area with similar skin texture and no skin problems to the scarred area was selected. In this embodiment, a scar-free area 1 cm in front of the ear or along the jawline could be selected. In the control area, 3-5 test points were also selected, and the same test procedures and parameters as those for the scarred area were used to measure and calculate the average value, which was then used as the baseline for the epidermal thickness of one's own healthy skin. Self-comparison can accurately reflect the thinness of the scarred epidermis and avoid comparison bias caused by differences in the thickness of healthy skin between different individuals.

[0057] In this embodiment, the skin barrier function index was obtained using a transepidermal water loss meter. The measured value directly reflects the integrity of the skin barrier; a higher transepidermal water loss meter value indicates poorer barrier function. The arithmetic mean of the transepidermal water loss meter values ​​at all test points in the scar area was calculated. The skin barrier function index was obtained by dividing 100 by the average of the transepidermal water loss meter values. The average skin barrier function index of healthy facial skin areas was also obtained by using healthy areas in front of the ear or along the jawline as controls. Transepidermal water loss meter measurements were performed at the corresponding test points, and the average value was calculated and converted into the barrier function index of the healthy area, thus visually reflecting the degree of barrier damage. In this embodiment, the relative thinness of the epidermis is calculated by dividing the epidermal thickness of the scar area by the average epidermal thickness of the healthy facial skin area. If the ratio is <1, it indicates that the epidermis of the scar area is thinner than the healthy skin; the smaller the ratio, the greater the thinness. The relative damage to the skin barrier is calculated by dividing the skin barrier function index of the scar area by the average barrier function index of the healthy area. If the ratio is <1, it indicates that the barrier function of the scar is worse than that of the healthy skin; the smaller the ratio, the more severe the damage. In this embodiment, since both the thinness of the epidermis and the degree of barrier damage are core factors affecting the risk of stress response, and the two are interrelated, a geometric mean is used to balance their weights, which can avoid the excessive influence of a single indicator on the results. Since the smaller the geometric mean result, the more severe the epidermal damage of the scar, and the higher the corresponding risk of stress response, its reciprocal is taken as the baseline for epidermal damage risk; the larger the baseline, the more severe the damage. In this embodiment, the risk of stress response in the scar area within a certain historical monitoring period is defined as the product of the baseline risk of epidermal damage and the proportion of the scar area. The baseline risk of epidermal damage reflects the vulnerability of the individual with the scar, while the proportion of the scar area reflects the extent of the impact. The product of the two can comprehensively reflect the risk of stress response in the scar area.

[0058] In this embodiment, step S3, which involves constructing a scar area stress response risk prediction model to predict the patient's scar area stress response risk in future environments, also includes the following specific steps:

[0059] S33. Based on the calculation process in step S31, obtain the environmental risk coefficients within multiple historical monitoring periods; based on the calculation process in step S32, obtain the scar area stress response risk within multiple historical monitoring periods; use the environmental risk coefficient within a certain historical monitoring period and the scar area stress response risk within the previous historical monitoring period corresponding to that historical monitoring period as the input dataset for that historical monitoring period; use the input datasets from multiple historical monitoring periods and the corresponding scar area stress response risks from multiple historical monitoring periods as the paired sample dataset; divide the paired sample dataset into a paired sample training set and a paired sample test set in a 7:3 ratio; in this embodiment, GBRT is used to analyze the complex correlation between environmental stimuli and scar stress response, providing algorithmic support for subsequent prediction. First, in this embodiment, the stress response risk in the scar area is jointly influenced by the environmental risk coefficient and historical stress risk. The relationship between the two is not simply linear (for example, stress risk increases slowly with environmental risk at low environmental risk and accelerates at high environmental risk). GBRT, through iterative ensemble of multiple decision trees, can effectively capture this nonlinear relationship and is robust to outliers, making it suitable for the small amount of noise that may exist in clinical monitoring data. Model training is implemented using the Python Scikit-learn library. The input features are the environmental risk coefficient and the stress response risk of the previous period in each period of the paired sample training set. The prediction target is the stress response risk of the scar area in the corresponding period in the training set. The model parameters need to be optimized through 5-fold cross-validation: the training set is randomly divided into 5 groups, each group is used as the validation set in turn, and the rest are used as the training set. By adjusting parameters such as the learning rate, the number of trees, and the maximum tree depth, the parameter combination with the smallest root mean square error of the validation set is selected. In this embodiment, the model training process employs gradient descent to minimize the loss function. Each iteration generates a decision tree, fitting the prediction residuals of the previous iteration to gradually reduce model error until the number of iterations reaches the required number of trees or the loss function converges. This embodiment utilizes the GBRT model, whose nonlinear fitting capability far surpasses linear regression, more realistically reflecting the correlation between the environment and scar stress. Cross-validation optimizes parameters to ensure model stability across different data subsets, avoiding prediction bias caused by inappropriate parameters and providing a reliable algorithmic model for accurate future stress risk prediction.

[0060] S34. Construct a gradient boosting regression tree model. Use the input dataset of all historical monitoring cycles in the paired sample training set as the input features of the gradient boosting regression tree model. Use the scar area stress response risk of all historical monitoring cycles in the paired sample training set as the prediction target of the gradient boosting regression tree model. Train the gradient boosting regression tree model to obtain the initial scar area stress response risk prediction model.

[0061] S35. The initial scar area stress response risk prediction model is tested using a paired sample test set to obtain the scar area stress response risk prediction model.

[0062] S36. Obtain the scar area stress response risk within the current monitoring period, and simultaneously extract the predicted UV index and air pollutant index for the next monitoring period from the environmental status data to calculate the environmental risk coefficient for the next monitoring period. Input the scar area stress response risk within the current monitoring period and the environmental risk coefficient for the next monitoring period into the scar area stress response risk prediction model for prediction. Use the predicted value output by the scar area stress response risk prediction model as the patient's scar area stress response risk in the future environment. In this embodiment, the scar area stress response risk prediction model, which has been trained and validated, is combined with the current scar status and future environmental forecast data to accurately predict the patient's scar stress response risk in the future environment. This provides data support for doctors to develop prospective intervention plans and helps patients avoid the adverse effects of the environment on scars in advance. First, in this embodiment, the scar area stress response risk for the current monitoring period is obtained according to step S32. Using the stress response risk for the current period as the input to the model can reflect the basic influence of the current physiological state of the scar on future stress. In this embodiment, the prediction of the environmental risk coefficient for the next monitoring cycle relies on authoritative environmental forecast data: the UV index forecast is obtained from the local meteorological department's website, with hourly UV index forecasts for the next 7 days, and the average value is taken as the UV index for the next cycle; the air pollutant index forecast is obtained from the local ecological and environmental bureau's website, and the average value is taken as the air pollutant index for the next cycle; the arithmetic average of the two is used to obtain the environmental risk coefficient for the next cycle. This embodiment uses the current cycle's stress response risk and the next cycle's environmental risk coefficient as input features, which are then substituted into a trained scar area stress response risk prediction model. The model's output prediction value is the patient's scar area stress response risk in the future environment. This embodiment enables prospective prediction of scar area stress response risk, breaking the limitations of traditional post-event assessments. It allows doctors and patients to understand scar stress risk 7-14 days in advance. For example, if the predicted future cycle stress risk reaches a high level, doctors can advise patients to adjust their outdoor plans, avoid going out during periods of strong UV radiation, and strengthen scar care. Combined with dynamic adjustments to the forecast data, the prediction is made more consistent with actual environmental changes, avoiding prediction bias caused by fixed environmental assumptions.

[0063] In this embodiment, step S4 involves constructing a dynamic evaluation model for scar recovery effectiveness to assess the risk of scar deterioration during the patient's scar repair process. This includes the following specific steps:

[0064] S41. Obtain the results of the intensity analysis of facial pollution behavior caused by the patient's daily behavior and the results of the prediction of the patient's risk of stress response in the scar area in the future environment;

[0065] S42. The comprehensive risk index is obtained by geometrically averaging the results of facial pollution behavior intensity analysis and the results of scar area stress response risk prediction.

[0066] S43. Extract the average epidermal repair rate and average collagen density of the scar area within the current monitoring period from the epidermal morphology data. Simultaneously, obtain the average epidermal repair rate and average collagen density of the healthy facial skin area within the current monitoring period. Divide the average epidermal repair rate of the scar area within the current monitoring period by the average epidermal repair rate of the healthy facial skin area within the current monitoring period to obtain the repair rate ratio. Divide the average collagen density of the scar area within the current monitoring period by the average collagen density of the healthy facial skin area within the current monitoring period to obtain the collagen density ratio. Perform a geometric mean calculation on the repair rate ratio and the collagen density ratio, and use the result of the geometric mean calculation as the patient's scar self-repair ability status. In this embodiment...

[0067] S44. Divide the comprehensive risk index by the patient's scar self-repair ability and use the result of the division as the risk of scar deterioration during the patient's scar repair process.

[0068] In this embodiment, step S5 involves issuing a medical follow-up warning to the patient based on the risk assessment results of scar recovery deterioration, including the following specific steps:

[0069] S51. Preset a medical follow-up warning threshold and obtain the risk assessment results of scar recovery and deterioration during the patient's scar repair process. In this embodiment, the process of setting the medical follow-up warning threshold is as follows: Collect monitoring data of 1000 patients with acne atrophic scars in the dermatology department of a certain hospital over the past 3 years, record the risk assessment results of scar recovery and deterioration for each patient and whether a follow-up label was triggered (a label value of 1 indicates that an emergency follow-up was actually performed due to scar deterioration, and a label value of 0 indicates that there was no deterioration during the regular follow-up). Determine the optimal warning threshold by plotting a receiver operating characteristic (ROC) curve: with the risk assessment results of scar recovery and deterioration as the abscissa and whether a follow-up label was triggered as the ordinate, calculate the sensitivity and specificity under different thresholds, and select the threshold with the largest area under the ROC curve and a balance between sensitivity and specificity as the medical follow-up warning threshold.

[0070] S52. Compare the risk assessment results of scar recovery deterioration during the patient's scar repair process with the medical follow-up warning threshold; if the risk assessment results of scar recovery deterioration during the patient's scar repair process are greater than the medical follow-up warning threshold, generate a medical follow-up warning signal for the patient; if the risk assessment results of scar recovery deterioration during the patient's scar repair process are less than or equal to the medical follow-up warning threshold, continue to monitor the patient's scar repair process.

[0071] Example 2

[0072] like Figure 2 As shown, this embodiment provides a computer vision-based facial acne atrophic scar assessment system, including: a data acquisition module, a facial pollution behavior intensity analysis module, a scar area stress response risk prediction module, a scar recovery and deterioration risk dynamic assessment module, and a medical follow-up early warning module. The data acquisition module acquires epidermal morphology data and scar location distribution data of the patient's facial acne atrophic scar area, as well as patient daily behavior data and environmental status data. The facial pollution behavior intensity analysis module, based on the patient's daily behavior data and scar location distribution data, constructs a facial pollution behavior intensity analysis model using computer vision technology to assess the facial pollution caused by the patient's daily behavior. The system analyzes the intensity of pollution behavior; the scar area stress response risk prediction module combines environmental status data with epidermal morphology data of the patient's facial acne atrophic scar area to construct a scar area stress response risk prediction model, predicting the patient's scar area stress response risk in future environments; the scar recovery deterioration risk dynamic assessment module, based on the facial pollution behavior intensity analysis results and scar area stress response risk prediction results, constructs a scar recovery deterioration risk dynamic assessment model, dynamically assessing the patient's scar recovery deterioration risk during the scar repair process; the medical follow-up early warning module, based on the dynamic assessment results of scar recovery deterioration risk, provides early warnings for patients requiring medical follow-up.

[0073] The parameters and steps for each unit module to achieve their respective functions in the computer vision-based facial acne pitted scar assessment system of the present invention described above can be referred to the parameters and steps in the embodiments of the computer vision-based facial acne pitted scar assessment method above, and will not be repeated here.

[0074] Example 3

[0075] An electronic device according to an embodiment of the present invention includes a processor and a memory, wherein the memory stores a computer program that can be called by the processor. The processor executes a computer vision-based method for assessing facial acne atrophic scars by calling the computer program stored in the memory. It should be noted that all computer programs for the computer vision-based method for assessing facial acne atrophic scars are implemented using C language. The data acquisition module, facial pollution behavior intensity analysis module, scar area stress response risk prediction module, scar recovery and deterioration risk dynamic assessment module, and medical follow-up early warning module are all controlled by a remote server.

[0076] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0077] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A computer vision based facial acne pitted scar assessment system, characterized in that, The system comprises a data acquisition module, a facial pollution behavior intensity analysis module, a scar area stress reaction risk prediction module, a scar recovery deterioration risk dynamic evaluation module, and a medical review warning module. The data acquisition module is configured to acquire epidermal morphology data and scar position distribution data of a facial acne pitted scar area of a patient, and simultaneously acquire daily behavior data and environmental state data of the patient. The facial pollution behavior intensity analysis module is configured to construct a facial pollution behavior intensity analysis model based on the daily behavior data and the scar position distribution data of the patient, and analyze facial pollution behavior intensity caused by daily behavior of the patient by using computer vision technology. The scar area stress reaction risk prediction module is configured to construct a scar area stress reaction risk prediction model by combining the environmental state data and the epidermal morphology data of the facial acne pitted scar area of the patient, and predict a scar area stress reaction risk of the patient in a future environment. The scar recovery deterioration risk dynamic evaluation module is configured to construct a scar recovery deterioration risk dynamic evaluation model based on the facial pollution behavior intensity analysis result and the scar area stress reaction risk prediction result, and dynamically evaluate a scar recovery deterioration risk of the patient in a scar repair process. The medical review warning module is configured to perform medical review warning on the patient according to the scar recovery deterioration risk dynamic evaluation result. The method for dynamically evaluating the scar recovery deterioration risk of the patient in the scar repair process by constructing the scar recovery deterioration risk dynamic evaluation model comprises the following specific steps: S41. Acquire facial pollution behavior intensity analysis results caused by daily behavior of the patient and scar area stress reaction risk prediction results of the patient in a future environment. S42. Perform geometric mean operation on the facial pollution behavior intensity analysis results and the scar area stress reaction risk prediction results to obtain a comprehensive risk index. S43. Analyze a repair rate ratio and a collagen density ratio based on the epidermal morphology data, and analyze a scar self-repairing ability state of the patient in combination with the repair rate ratio and the collagen density ratio. S44. Perform division operation on the comprehensive risk index and the scar self-repairing ability state of the patient, and take a result of the division operation as the scar recovery deterioration risk of the patient in the scar repair process.

2. The computer vision based facial acne depression scar assessment system of claim 1, wherein, The method for analyzing facial pollution behavior intensity caused by daily behavior of the patient by constructing the facial pollution behavior intensity analysis model based on computer vision technology comprises the following specific steps: S21. Extract a total number of times of hand contact with the face and a duration of each contact in a current monitoring period from the daily behavior data of the patient, take a ratio of the total number of times of hand contact with the face to the number of monitoring days as a daily contact frequency, and take an average value of the duration of each contact as an average single contact duration; and perform multiplication operation on the daily contact frequency and the average single contact duration to obtain a behavior contact intensity. S22. Extract a total area of the scar area from the scar position distribution data, and extract an overlapping area of the contact area and the scar area in the current monitoring period from the daily behavior data. The ratio of the overlapping area to the total hand contact area in the monitoring period is taken as a scar area contact proportion; the ratio of the overlapping area to the total scar area is taken as a scar area coverage proportion; the scar area contact proportion and the scar area coverage proportion are arithmetically averaged to obtain a position correlation strength; S23, the behavior contact strength and the position correlation strength are geometrically averaged, and the geometrically averaged result is taken as a facial pollution behavior strength caused by the patient's daily behavior.

3. The computer vision based facial acne scar depression assessment system of claim 2, wherein, The scar area stress reaction risk prediction model is constructed to predict the scar area stress reaction risk of the patient in a future environment, and includes the following specific steps: S31, the ultraviolet index and the air pollutant index in a certain historical monitoring period are extracted from the environmental state data; the ultraviolet index and the air pollutant index are arithmetically averaged to obtain an environmental risk coefficient in the certain historical monitoring period; S32, the epidermis thickness value and the skin barrier function index of the patient's scar area in the certain historical monitoring period are extracted from the epidermis morphology data; the epidermis thickness average value and the skin barrier function index average value of the patient's facial healthy skin area in the certain historical monitoring period are obtained; the epidermis thickness value of the patient's scar area is divided by the epidermis thickness average value of the patient's facial healthy skin area to obtain an epidermis relative weakness degree; the skin barrier function index of the patient's scar area is divided by the skin barrier function index average value of the patient's facial healthy skin area to obtain a barrier relative damage degree; the epidermis relative weakness degree and the barrier relative damage degree are geometrically averaged, and the reciprocal of the geometrically averaged result is taken as an epidermis damage danger base; the ratio of the total scar area to the total facial area is taken as a scar area proportion from the scar position distribution data; the epidermis damage danger base and the scar area proportion are multiplied to obtain a scar area stress reaction risk in the certain historical monitoring period.

4. The computer vision based facial acne scar depression assessment system of claim 3, wherein, The scar area stress reaction risk prediction model is constructed to predict the scar area stress reaction risk of the patient in a future environment, and includes the following specific steps: S33, based on the calculation process of step S31, the environmental risk coefficients in multiple historical monitoring periods are obtained; based on the calculation process of step S32, the scar area stress reaction risks in multiple historical monitoring periods are obtained; the environmental risk coefficient in a certain historical monitoring period and the scar area stress reaction risk in the corresponding previous historical monitoring period in the certain historical monitoring period are taken as an input data set of the certain historical monitoring period, the input data sets of multiple historical monitoring periods and the scar area stress reaction risks of the corresponding multiple historical monitoring periods are taken as a paired sample data set, and the paired sample data set is divided into a paired sample training set and a paired sample test set in a ratio of 7:3; S34, a gradient boosting regression tree model is constructed, all input data sets of the historical monitoring periods in the paired sample training set are taken as input features of the gradient boosting regression tree model, all scar area stress reaction risks of the corresponding historical monitoring periods in the paired sample training set are taken as prediction targets of the gradient boosting regression tree model, the gradient boosting regression tree model is trained, and an initial scar area stress reaction risk prediction model is obtained; S35, the initial scar area stress reaction risk prediction model is tested by using the paired sample test set, and a scar area stress reaction risk prediction model is obtained; S36, a scar area stress reaction risk in a current monitoring period is obtained, an ultraviolet index and an air pollutant index in a next monitoring period predicted in the environmental state data are extracted, and an environmental risk coefficient in the next monitoring period is calculated; the scar area stress reaction risk in the current monitoring period and the environmental risk coefficient in the next monitoring period are input into the scar area stress reaction risk prediction model for prediction, and a prediction value output by the scar area stress reaction risk prediction model is taken as a scar area stress reaction risk of the patient in a future environment.

5. The computer vision based facial acne scar depression assessment system of claim 4, wherein, The medical review warning of the patient is performed according to the dynamic evaluation result of the scar recovery deterioration risk, and the medical review warning of the patient includes the following specific steps: S51, a medical review warning threshold is preset, and a scar recovery deterioration risk evaluation result in a scar repair process of the patient is obtained; S52, the scar recovery deterioration risk evaluation result in the scar repair process of the patient is compared with the medical review warning threshold; If the scar recovery deterioration risk evaluation result in the scar repair process of the patient is greater than the medical review warning threshold, a medical review warning signal of the patient is generated; If the scar recovery deterioration risk evaluation result in the scar repair process of the patient is less than or equal to the medical review warning threshold, the scar repair process of the patient is continuously monitored.

6. A computer vision-based method for facial acne atrophic scarring assessment, implemented based on the computer vision-based system for facial acne atrophic scarring assessment of any one of claims 1-5, characterized in that, The following specific steps are included: The epidermal morphology data and the scar position distribution data of the facial acne pitted scar area of the patient are obtained, and the daily behavior data and the environmental state data of the patient are obtained; Based on the daily behavior data and the scar position distribution data of the patient, a facial pollution behavior intensity analysis model is constructed by using computer vision technology, and the facial pollution behavior intensity caused by the daily behavior of the patient is analyzed; Combined with the environmental state data and the epidermal morphology data of the facial acne pitted scar area of the patient, a scar area stress reaction risk prediction model is constructed, and the scar area stress reaction risk of the patient in a future environment is predicted; Based on the facial pollution behavior intensity analysis result and the scar area stress reaction risk prediction result, a scar recovery deterioration risk dynamic evaluation model is constructed, and the scar recovery deterioration risk in the scar repair process of the patient is evaluated; The medical review warning of the patient is performed according to the dynamic evaluation result of the scar recovery deterioration risk.

7. An electronic device comprising: A processor and a memory, wherein the memory stores a computer program that can be called by the processor; characterized in that the processor executes the computer vision-based facial acne pitted scar evaluation method of claim 6 by calling the computer program stored in the memory.

Images