Information planning system for patients with acne based on IMB model

Abstract

This invention relates to the field of intelligent medical information technology, providing an information-based planning system for acne patients based on the IMB model. The system collects facial acne characteristics, medication adherence, and facial oil characteristics data from acne patients. It constructs a gridded oil-matching medication adherence index and a gridded acne condition index. Through oil secretion gradient weighted fusion and dynamic condition calibration, it outputs the facial gridded drug utilization rate results. Furthermore, it identifies the influencing factors of each facial grid and performs spatial aggregation to generate medication adherence planning instructions that match the individual's facial region characteristics. This invention forms a closed-loop process from condition gridded perception, multi-dimensional behavior quantification, drug utilization rate attribution assessment to regional targeted intervention. It achieves refined and intelligent acne medication management based on differences in facial oil secretion and lesion aggregation characteristics, significantly improving the accuracy of individualized intervention for acne patients' topical medication behavior.

Description

Technical Field

[0001] This invention belongs to the field of intelligent medical information technology, specifically relating to an information planning system for acne patients based on the IMB model. Background Technology

[0002] The clinical efficacy of topical acne medication treatment highly depends on patient adherence and accuracy of administration. Existing medication behavior management systems are mainly based on fixed-time reminders or general medication check-in mechanisms. Their core logic remains at the binary recording level of medication behavior, which has two technical limitations: First, existing technologies do not fully consider the significant differences in sebum secretion rates in different facial areas on drug absorption, adhesion, and efficacy. The sebum secretion rate in different areas of the same patient's face can differ by 2-4 times, and the difference between patients can be as high as 30 times. This key individual characteristic directly determines the effective concentration and duration of action of topical drugs on the skin surface, but existing systems can only perform rough skin type classification of oily or dry, and cannot achieve precise quantification of regional sebum secretion characteristics on the face.

[0003] Secondly, existing technologies only record whether or not medication is used in a binary manner. They cannot determine whether the timing of medication matches the drug's photostability or other properties, nor can they accurately correct medication administration behavior by combining the individual facial sebum secretion characteristics of the patient. As a result, the clinical effectiveness of the drug cannot be quantitatively evaluated, and medication intervention strategies remain at the post-event feedback stage.

[0004] Therefore, there is an urgent need for an information-based planning system for acne patients' medication behavior that integrates multi-dimensional data on regional facial sebum secretion characteristics and medication adherence behavior. Summary of the Invention

[0005] In order to overcome the defects and shortcomings of existing technologies, this invention provides an information planning system for acne patients based on the IMB model.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] In a first aspect, the present invention provides an information planning system for acne patients based on the IMB model, comprising the following modules:

[0008] Data acquisition module: Simultaneously collects facial acne characteristic data, facial medication administration data, and facial oil characteristic data of acne patients;

[0009] Feature extraction module: Based on facial oil feature data and facial medication execution data, generate oil-matching medication execution index; based on acne lesion type and spatial clustering features in facial acne feature data, generate grid acne condition index.

[0010] Behavioral assessment module: Based on facial oil feature data, it integrates and analyzes the grid acne disease index and the oil-matching medication execution index to output the facial grid-based drug utilization rate results;

[0011] Intervention generation module: Based on the results of facial grid-based drug utilization, it generates medication execution planning instructions that match the individual facial region characteristics of acne patients.

[0012] According to the above technical solution, the steps of simultaneously collecting facial acne characteristic data, facial medication administration data, and facial oil characteristic data of acne patients include:

[0013] S110. Collect facial image data of acne patients using image recognition algorithms, and construct facial acne feature data including acne lesion type, acne lesion count, and acne lesion location coordinates, specifically including:

[0014] Acne patients' facial image data is acquired at a preset period. The facial area is divided into facial grids of preset size. The acquired facial image data is processed using an image recognition algorithm to identify and extract the acne lesion type and distribution information of each acne lesion type in each facial grid. The acne lesion type, acne lesion count, and location coordinates of each acne lesion are generated for each facial grid at the preset period acquisition time. The acne lesion type, count, and location coordinates are combined in the order of acquisition time to form a facial acne feature data sequence with time tags.

[0015] S120. Based on acne patient medication events, generate facial medication execution data including medication timestamps, application areas, acne medication identifiers, and acne medication application dosages, specifically including:

[0016] Medication execution data for acne patients is collected for each medication event. The medication timestamp is generated when the acne patient completes the medication. The application area is generated by the acne patient selecting one or more facial grids to be covered in this medication in a preset standardized facial grid interface. The drug identifier is obtained by scanning the barcode of the acne patient's topical acne medication.

[0017] The dosage is obtained by weighing the same tube of ointment before and after a single application of acne medication to determine the total dosage for this application. Based on the number of selected application grids, the total dosage is allocated to each covered facial grid to obtain the acne medication dosage for each covered facial grid.

[0018] The medication timestamps, application areas, acne drug identifiers, and acne drug application amounts corresponding to each medication event are combined in chronological order to form a medication execution data sequence with time tags;

[0019] The medication execution data sequence is associated with the facial acne feature data sequence collected at a preset period through time tags. The medication event data within the period unit is aggregated using the time of two adjacent facial image data collections as the period unit, and is used to analyze the impact of medication behavior on changes in acne lesions within the preset period.

[0020] S130. Collect facial sebum secretion rate data from acne patients, and construct facial sebum characteristic data to characterize the sebum secretion features of each facial grid, specifically including:

[0021] After cleansing the facial skin, two time-series measurements of sebum volume were performed on the core sebum detection functional areas of the face. Based on the difference and interval between the two measurements, the measured sebum secretion rate of each core sebum detection functional area was calculated. Based on the anatomical distribution of facial sebum secretion, the measured sebum secretion rate of each core sebum detection functional area was mapped and fitted to each facial grid within the corresponding coverage area to obtain the sebum secretion rate value corresponding to each facial grid, forming full-face gridded facial oil feature data.

[0022] According to the above technical solution, the steps of generating an oil-matching medication execution index based on facial oil feature data and facial medication execution data, and generating a grid acne condition index based on acne lesion types and spatial clustering characteristics in facial acne feature data, include:

[0023] S210. Based on the facial oil feature data of acne patients, obtain the oil secretion gradient features corresponding to each facial grid of acne patients, specifically including:

[0024] The sebum secretion rate values ​​of each facial grid are extracted from the facial sebum feature data; according to the preset sebum secretion grading rules, the sebum secretion rate value of each facial grid is mapped to the corresponding sebum secretion gradient feature, which is used to characterize the sebum secretion level of each facial grid.

[0025] S220. Based on the sebum secretion gradient characteristics of each facial grid, a matching analysis is performed on the medication timestamp, application area, and acne medication application amount to obtain an oil-matching medication execution index that includes oil-matching time compliance, oil-matching spatial overlap, and oil-matching dosage compliance. Specifically, it includes:

[0026] Based on facial medication execution data, the acne drug identification is obtained from a pre-set acne topical drug knowledge base. The general usage guidelines for the drug include recommended usage time, applicable skin type, and standard dosage range per unit area.

[0027] Combining the sebum secretion gradient characteristics of each facial grid with the general drug usage guidelines, a sebum matching execution benchmark is determined for each facial grid. The sebum matching execution benchmark is dynamically adjusted based on the sebum secretion gradient to meet the general drug usage guidelines. The sebum matching execution benchmark is used to characterize the optimal execution standard for topical acne drugs under different facial sebum secretion states, including the time execution benchmark, spatial coverage benchmark, and dosage execution benchmark that match the sebum secretion gradient of the grid.

[0028] Extract the medication timestamps corresponding to each medication event, the set of covered facial grids corresponding to the application area, and the amount of acne medication applied to each covered facial grid from the facial medication execution data. Combine this with the standardized area of ​​the facial grid to calculate the actual amount of medication per unit area for each facial grid.

[0029] The oil matching time compliance of the facial grid is obtained by comparing the medication timestamp in the facial medication execution data with the time execution benchmark of the facial grid.

[0030] The S230 pre-divided skin lesion cluster area is obtained from the facial acne feature data. The spatial overlap between the facial mesh corresponding to the application area and the skin lesion cluster area is calculated. The spatial coverage benchmark corresponding to the sebum secretion gradient feature of the facial mesh is adjusted to obtain the sebum matching spatial overlap of the facial mesh.

[0031] The actual amount of medication per unit area for each facial grid is compared with the dosage execution benchmark for that facial grid to obtain the compliance of the oil-matching dosage for that facial grid.

[0032] S230. Based on the dynamic trends of acne lesion types and numbers in acne patients, adjusted disease severity weights are assigned to each type of acne lesion according to differences in clinical severity, specifically including:

[0033] The location coordinates of each acne lesion in the current period are extracted from the facial acne feature data. Based on the spatial proximity relationship, the acne lesions are divided into several lesion cluster areas. Each lesion cluster area corresponds to a continuous dense area of ​​lesions on the face.

[0034] For each lesion cluster area, the number of acne lesions of each type in the lesion cluster area is counted. According to the clinical severity classification, the static basic condition weight is preset for each type of lesion. The higher the severity, the greater the weight. The total static basic condition weight of the lesion cluster area is obtained by summing them up.

[0035] The location coordinates of each acne lesion in the previous period are extracted from the facial acne feature data. The lesion cluster area in the current period is spatially matched with that in the previous period. The total number of acne lesions in the same area in adjacent periods is compared to determine the trend of disease change: an increase in the number of lesions indicates disease progression, a decrease in the number of lesions indicates lesion regression, and no change in the number of lesions indicates disease stability. A correction coefficient for the number of lesions is assigned to different trends. The correction coefficient for the number of lesions is increased for disease progression, decreased for lesion regression, and 1 for disease stability.

[0036] The static baseline disease weight of each skin lesion cluster area is multiplied by the corresponding quantity change correction coefficient to obtain the corrected disease weight of that skin lesion cluster area, which is used to characterize the overall disease severity of that skin lesion cluster area in the current cycle.

[0037] S240. Based on the spatial distribution of different types of acne lesions on the face of acne patients, calculate the degree of lesion clustering in the region and generate a regional clustering coefficient; fuse the corrected disease weights with the regional clustering coefficients to generate a grid acne disease index, specifically including:

[0038] The density of acne lesions within the current period's lesion cluster area is calculated to generate a regional clustering coefficient. This regional clustering coefficient is used to characterize the concentration risk of acne lesions within the lesion cluster area. The denser the acne lesions, the higher the regional clustering coefficient.

[0039] The adjusted disease weight of each skin lesion cluster area is fused with the corresponding regional clustering coefficient to obtain the comprehensive disease risk value of the skin lesion cluster area;

[0040] The comprehensive disease risk value of each skin lesion cluster area is assigned to the corresponding facial grid according to the spatial coverage relationship between each facial grid and the skin lesion cluster area, generating the grid acne disease index of each facial grid, which is used to characterize the disease severity and cluster risk of the skin lesion cluster area in each facial grid.

[0041] According to the above technical solution, the step of fusing and analyzing the acne condition index and the oil-matching medication execution index based on facial oil feature data to output the facial grid-based drug utilization rate result includes:

[0042] S310. Based on the sebum secretion gradient characteristics of each facial grid in acne patients, the sebum-matched medication execution index is weighted and fused, and dynamically calibrated in conjunction with the grid acne condition index to calculate the drug utilization rate score for each facial grid, specifically including:

[0043] A weight allocation scheme is preset for the different sebum secretion gradient features of each facial grid. The weight allocation scheme is based on three indicators: sebum matching time compliance, sebum matching spatial overlap, and sebum matching dosage compliance. These indicators are used to characterize the differences in the contribution of the execution quality indicators of each medication behavior to the drug utilization rate under different sebum secretion states.

[0044] Based on the weight allocation scheme corresponding to the sebum secretion gradient feature of the facial mesh, the three indicators of sebum matching time compliance, sebum matching spatial overlap and sebum matching dosage compliance are weighted and summed to calculate the basic behavior score of the facial mesh. The basic behavior score reflects the theoretical contribution of the quality of drug use behavior execution to drug utilization rate under the sebum state.

[0045] Based on the acne condition index of the facial grid, the condition penetration loss coefficient is calculated, and the basic behavioral score is dynamically calibrated to calculate the drug utilization rate score. The calibration rule is as follows: the higher the acne condition index, the greater the difficulty of drug penetration and action in the lesion area; the lower the condition penetration loss coefficient, the greater the downward adjustment of the basic behavioral score, which is used to objectively reflect the limiting effect of the severity of the lesion on the actual drug utilization rate.

[0046] The drug utilization score is used to quantitatively characterize the expected level of clinically effective drug utilization, determined by the quality of actual drug administration behavior in the facial grid, under the current sebum secretion state and disease background.

[0047] S320. Based on the drug utilization rate scores of each facial grid, classify the drug utilization rate levels, and based on the decomposition results of disease impairment loss and behavioral deviation loss, identify the core influencing factors leading to insufficient drug utilization rate of the facial grid, specifically including:

[0048] The drug utilization rate score of each facial grid is compared with the preset score threshold, and two utilization rate levels, high and low, are used as the drug utilization rate level of the facial grid.

[0049] For a target facial mesh with a low drug utilization level, the total loss value of the drug utilization score is calculated, where the total loss value is the difference between the full score of the drug utilization score and the drug utilization score.

[0050] To accurately distinguish the core sources of drug utilization loss, the total loss value is divided into two independent parts: disease severity loss and behavioral deviation loss. Disease severity loss reflects the objective limiting effect of lesion severity on drug penetration and utilization, while behavioral deviation loss reflects the drug utilization loss caused by patients' medication behavior deviating from the lipid matching execution benchmark.

[0051] A preset threshold for determining the dominant loss type is established. Based on the proportion of two types of losses—illness-related loss and behavioral deviation loss—in the total loss value, the dominant loss type of the target facial mesh is determined: if the proportion of illness-related loss exceeds the preset threshold, the dominant loss type is determined to be illness-related; otherwise, the dominant loss type is determined to be behavioral-related.

[0052] For target facial meshes with different loss-dominant types, corresponding core influencing factor identification rules are matched: For behavior-dominant facial meshes, the single-dimensional contribution of three oil-matching medication execution indices to behavioral deviation loss is calculated, and the dimension with the largest contribution is determined as the core influencing factor; For disease-dominant facial meshes, the severity of lesions and the difficulty of drug penetration are determined as the core influencing factors.

[0053] The drug utilization rate level, loss dominance type, and core influencing factors of each facial grid are uniformly used as the output of the behavior assessment module and transmitted to the subsequent intervention generation module.

[0054] According to the above technical solution, the step of generating medication execution planning instructions that match the individual facial region characteristics of acne patients based on facial gridded drug utilization results includes:

[0055] S410. Aggregate spatially adjacent facial grids with consistent core influencing factors to form clinically interventionable facial region units, specifically including:

[0056] The pre-defined rule for determining the spatial adjacency of standardized facial grids is the 4-neighborhood rule, where edges are directly adjacent. Using the core influence factor as the core aggregation condition, and combined with the drug utilization rate level, target facial grids that simultaneously meet the criteria of spatial adjacency, belong to the low drug utilization rate level, and have the same core influence factor are clustered and merged. For isolated target facial grids with low drug utilization rate that cannot be merged, they are assigned to adjacent intervention areas of the same type according to the core influence factor.

[0057] Ultimately, several continuous, clinically operable facial intervention regions are formed. Each facial intervention region consists of multiple spatially continuous low-utilization standardized facial grids, possessing unified core influencing factors, unified drug utilization echelons, and corresponding to a unique clinical intervention target.

[0058] S420. Based on the drug utilization rate level and core influencing factors of each facial region unit, match the corresponding drug administration execution plan instructions for each facial region unit, specifically including:

[0059] Based on the core influencing factor types of each facial intervention area unit, corresponding basic intervention strategies are matched, covering two core influencing scenarios: medication behavior deviation and lesion condition constraints. The matched basic intervention strategies are then combined with the facial anatomical coverage and drug utilization rate level of the facial intervention area unit to transform into medication execution planning instructions that acne patients can intuitively understand. The medication execution planning instructions for each facial area are summarized, and basic care and follow-up visit requirements are supplemented based on the overall condition of the acne patient's face, forming and outputting the overall personalized medication execution plan for the current cycle unit.

[0060] Secondly, this application provides an electronic device, including: a processor and a memory, wherein the memory stores a computer program that can be called by the processor, and the processor executes an acne patient information planning system based on the IMB model by calling the computer program stored in the memory.

[0061] Thirdly, this application provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to execute an information planning system for acne patients based on the IMB model.

[0062] Compared with the prior art, this application has the following advantages and beneficial effects:

[0063] This application establishes personalized facial medication guidelines that match local skin oil status and acne condition by taking the facial grid-like sebum secretion characteristics as the core. It achieves refined quantification of medication behavior through multi-dimensional medication execution index; it integrates the risk of acne lesion clustering to construct a grid-like condition index, and achieves accurate assessment of drug utilization through oil-differentiated weighting and dynamic calibration of condition, and locates core influencing factors based on loss decomposition.

[0064] Ultimately, by generating clinically interventionable areas through grid-based targeted aggregation, and matching targeted intervention strategies, a closed-loop process from data collection, assessment and diagnosis to targeted intervention is formed. This solves the shortcomings of existing technologies, such as uniform medication for the entire face, extensive management, and lack of precise root cause localization, and significantly improves the individualization level and intervention accuracy of acne outpatient medication management. Attached Figure Description

[0065] Figure 1 This is a schematic diagram of the structure of an information planning system for acne patients based on the IMB model provided in this application embodiment;

[0066] Figure 2 This is a flowchart illustrating the feature extraction module provided in an embodiment of this application;

[0067] Figure 3 This is a flowchart illustrating the behavior assessment module provided in an embodiment of this application;

[0068] Figure 4 This is a flowchart illustrating the intervention generation module provided in an embodiment of this application. Detailed Implementation

[0069] The technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments and specific features in the embodiments are detailed descriptions of the technical solution of this application, rather than limitations thereof. In the absence of conflict, the embodiments and technical features in the embodiments can be combined with each other.

[0070] Please see Figure 1 , Figure 1 This is a schematic diagram of the structure of an information planning system for acne patients based on the IMB model provided in this application embodiment, which specifically includes the following modules:

[0071] Data acquisition module: Simultaneously collects facial acne characteristic data, facial medication administration data, and facial oil characteristic data of acne patients.

[0072] In this embodiment, the simultaneous collection of facial acne characteristic data, facial medication administration data, and facial oil characteristic data of acne patients includes the following specific contents:

[0073] S110. Collect facial image data of acne patients through image recognition algorithms, and construct facial acne feature data including acne lesion type, acne lesion count, and acne lesion location coordinates.

[0074] At a preset weekly cycle (which matches the natural evolution cycle of acne lesions, effectively capturing dynamic changes in the condition while ensuring the continuity of data collection and patient compliance), standardized image acquisition equipment (such as smartphones or medical-grade digital cameras equipped with high-resolution cameras) is used to capture facial images of acne patients under uniform lighting calibration conditions. During the capture, patients are required to keep their faces clean and unobstructed, and to take frontal and 45-degree side views to cover the entire facial skin area.

[0075] The collected raw facial image data of acne patients underwent preprocessing, including size normalization, color correction, and facial region cropping, to form a unified format of facial image samples. Specifically, the following steps were performed: First, all raw facial images of acne patients were normalized, scaling them to a preset pixel size of 1024×1024 pixels to ensure a consistent scale across images taken from different devices and distances. Then, color correction was performed, adjusting the RGB channels of the raw facial images of acne patients based on a standard color chart or a facial skin reference area. Color space conversion and color temperature correction are performed to eliminate color deviations caused by differences in shooting environments and equipment. Finally, facial region cropping is performed. Based on facial key point detection algorithms (such as stacked hourglass networks or HRNet, which are pre-trained on a large-scale face dataset containing 68 facial key point annotations and can locate the precise pixel coordinates of key points such as the inner corner of the eye, the tip of the nose, and the corner of the mouth), feature points such as the inner corner of the eye, the tip of the nose, and the corner of the mouth are located. The minimum bounding rectangle region of the face is delineated based on the coordinates of the feature points, and non-facial areas such as the background, hair, and neck are cropped to form a facial image sample containing only the facial skin.

[0076] Based on the aforementioned calibrated facial key points, a unified facial coordinate system is established. Under this coordinate system, the facial region is divided into several standardized facial grids of preset sizes, completing the grid mapping of the entire facial region and providing a unified spatial benchmark for subsequent grid-level skin lesion feature extraction and analysis. In this embodiment, the preset size is set to a 10mm×10mm square grid by default. The setting is based on the common diameter of 1-3mm for single acne skin lesions, the accuracy of image recognition algorithms, and the practicality of clinical medication guidance. It can be flexibly adjusted according to actual needs.

[0077] A deep convolutional neural network-based image recognition algorithm was used to process facial image data collected at each preset period. The image recognition algorithm was a semantic segmentation model pre-trained on a medical image dataset labeled with various types of acne lesions. This semantic segmentation model adopted an encoder-decoder architecture (such as U-Net or DeepLab series). The encoder part gradually extracted multi-scale depth features of the image through stacked convolutional layers and pooling layers, and the decoder part restored the original image resolution through upsampling and skip connections. Finally, it output a probability map of each pixel belonging to various types of acne lesions. The model was pre-trained on an acne lesion dataset. Each image in the acne lesion dataset was labeled at the pixel level, and the labeling content covered three types of lesion regions: non-inflammatory comedones, inflammatory papules and pustules, and nodules and cysts. The model parameters were optimized by using the cross-entropy loss function to achieve the segmentation accuracy required for clinical applications.

[0078] For each input facial image sample, the semantic segmentation model outputs a probability prediction map of three types of acne lesions. After segmentation using a conditional random field or thresholding, a binarized lesion segmentation mask is obtained, where each pixel is assigned a category label of one of the following: non-inflammatory comedones, inflammatory papules / pustules, or nodules / cysts. Based on the lesion segmentation mask, a connected component analysis algorithm is used to extract each independent lesion region according to the category label: the acne lesion type corresponding to each independent lesion region is recorded; the centroid pixel coordinates of each independent lesion region are calculated, and affine transformation parameters are calculated based on the correspondence of facial key points. The centroid pixel coordinates are uniquely mapped to a unified facial coordinate system (this facial coordinate system is calibrated based on the facial key point detection algorithm to locate feature points such as the inner corner of the eye, the tip of the nose, and the corner of the mouth, and the same facial position in facial images at different acquisition times is mapped to a unified coordinate system through affine transformation) to obtain the acne lesion location coordinates.

[0079] Based on the location coordinates of each acne lesion, the corresponding facial grid to which the acne lesion belongs is matched to complete the spatial association between the single lesion and the facial grid; the acne lesion type, acne lesion count, and relative position coordinates of each acne lesion in the facial grid are counted for each facial grid; the number of each type of acne lesion area is counted to obtain the total number of acne lesions on the face.

[0080] The acne lesion type, acne lesion count, location coordinates, and associated facial grid information of each facial grid are combined in chronological order of collection time to form a facial acne feature data sequence with time labels and grid-level dimensions. This facial acne feature data sequence records the dynamic evolution of acne lesions in different facial grids at different time points, providing basic data support for subsequent quantification of disease condition and analysis of medication behavior.

[0081] S120. Based on medication events of acne patients, generate facial medication execution data including medication timestamps, application areas, acne drug identifiers, and acne drug application dosages.

[0082] After each application of topical medication, acne patients record the current time to generate a medication timestamp; the acne medication identifier is obtained by scanning the product barcode or electronic supervision code of the topical acne medication used by the acne patient in this application.

[0083] Meanwhile, acne patients confirm the application area in the preset standard facial grid interface. The standard facial grid interface and the unified facial coordinate system in S110 are the same spatial reference system. Based on the facial key point detection algorithm, feature points such as the inner corner of the eye, the tip of the nose, and the corner of the mouth are calibrated and constructed, which correspond one-to-one with the standardized facial grids divided in S110, ensuring that the application area and the coordinates of the acne lesion are accurately matched in the same spatial reference system. Acne patients select one or more facial grids to be covered by this medication, record the unique identifier and coordinate range of all selected facial grids, and generate the application area for this medication.

[0084] The dosage of acne medication was obtained using a high-precision electronic balance (accuracy 0.001g) that had been calibrated for measurement: Before and after each application of the topical acne medication, the same tube of ointment was weighed by the acne patient, and the difference between the two weighing results was the total dosage for this application; based on the total number of selected application grids, the total dosage for this application was divided equally according to the number of grids to obtain the dosage of acne medication for each grid on the covered face.

[0085] The medication timestamp, application area, acne drug identifier, and acne drug application amount corresponding to each facial grid are combined in chronological order to form a medication execution data sequence with time tags. The medication execution data sequence is associated with the facial acne feature data sequence collected according to a preset period through time tags. Taking the time of two adjacent facial image acquisitions as the period unit, all medication event data within the period unit are aggregated to analyze the impact of medication behavior on changes in acne lesions within the preset period.

[0086] S130. Collect facial sebum secretion rate data from acne patients and construct facial sebum characteristic data to characterize the sebum secretion features of each facial grid.

[0087] First, the acne patient cleanses their entire face to remove residual oil, skincare products, makeup, sweat, and other interfering substances. Then, in a normal room temperature and humidity environment (avoiding extreme high temperature, high humidity, or strong wind environments, matching the daily skin condition), the facial skin is allowed to rest for a preset time (generally no less than 15 minutes to ensure that sebum secretion returns to stability) to eliminate temporary interference from environmental factors, strenuous exercise, emotional fluctuations, etc., and to ensure the objectivity and repeatability of the test data.

[0088] Based on the standardized facial grid defined in S110, and the anatomical distribution of facial sebum secretion, four core sebum detection functional areas are pre-defined: the central area of ​​the forehead, the central area of ​​the nose, the bilateral cheek areas, and the central area of ​​the jaw. Each core sebum detection functional area corresponds to a continuous facial grid within a fixed range.

[0089] A calibrated skin sebum meter was used to measure sebum volume twice in each core sebum detection functional area. The two measurements were recorded after an interval of 2-4 hours. The measured sebum secretion rate of each core sebum detection functional area was calculated by the ratio of (second measurement value - first measurement value) to the interval.

[0090] Based on the inherent anatomical gradient of sebum secretion—highest in the central forehead and nose, decreasing linearly towards the cheeks, jawline, and facial edges—the measured sebum secretion rates of the four core sebum detection zones were mapped and fitted to a standardized full-face grid defined by S110 according to the following fixed rules, resulting in a unique sebum secretion rate value for each facial grid:

[0091] The core sebum detection function area directly covers the facial grid: all facial grids within the range of the four core sebum detection function areas are directly assigned the measured value of the sebum secretion rate of the corresponding core sebum detection function area, which is used as the facial sebum secretion rate data of that facial grid.

[0092] The transitional facial grid between core sebum detection functional areas: The facial grid located between two adjacent core sebum detection functional areas uses a two-point linear interpolation method to calculate the fitted value. The fixed calculation formula is: Facial sebum secretion rate value of the grid to be fitted = measured value of the high-value core area × (1 - distance ratio) + measured value of the low-value core area × distance ratio; where distance ratio = number of grids from the facial grid to the high-value core area ÷ total number of facial grids between the two core detection functional areas; In addition, the higher measured value of sebum secretion rate between two adjacent core sebum detection functional areas is defined as the measured value of the high-value core area, and the lower value is defined as the measured value of the low-value core area, strictly following the law that sebum secretion gradually decreases from the high-value area to the low-value area;

[0093] Facial edge grid: Facial edge grids that are not covered by the core sebum detection functional area and have no adjacent core sebum detection functional areas are directly matched with the facial sebum secretion rate data of the nearest facial grid as the facial sebum secretion rate data of that facial edge grid.

[0094] Feature extraction module: Based on facial oil feature data and facial medication execution data, it generates an oil-matching medication execution index; based on acne lesion types and spatial clustering features in facial acne feature data, it generates a grid acne condition index.

[0095] In this embodiment, as Figure 2 As shown, based on facial oil feature data and facial medication execution data, an oil-matching medication execution index is generated. Based on acne lesion types and spatial clustering characteristics in facial acne feature data, a grid acne condition index is generated, including the following specific steps:

[0096] S210. Based on the facial oil feature data of acne patients, obtain the oil secretion gradient features corresponding to each facial grid of acne patients.

[0097] Extract facial sebum secretion rate data corresponding to each facial grid, and map the sebum secretion rate value of each facial grid to the corresponding sebum secretion gradient feature according to the preset sebum secretion grading rules. The sebum secretion gradient feature is used to quantitatively characterize the sebum secretion level and acne incidence risk level of each facial grid.

[0098] The sebum secretion grading rule combines the normal physiological values ​​of facial sebum secretion and the clinical threshold for acne onset, dividing it into two fixed gradient levels based on the sebum secretion rate: when the facial mesh sebum secretion rate value is < When the facial mesh has a low sebum secretion gradient, and the sebum secretion rate of the facial mesh is ≥ At that time, the facial mesh was a high oil secretion gradient;

[0099] The core basis for this sebum secretion grading rule is: firstly, it conforms to clinical skin physiology benchmarks. The following are the clinically recognized normal physiological ranges for facial sebum secretion in healthy individuals in China, which conform to the basic physiological characteristics of human skin and are universally applicable to the entire population; secondly, they match the clinical and pathological thresholds for acne development, and existing clinical evidence data have confirmed that a consistently higher sebum secretion rate than [previous range] indicates [a certain condition]. The risk of acne development increases significantly at this time. This threshold can accurately distinguish between low-risk and high-risk facial areas for acne, providing a clear grading benchmark for the subsequent development of differentiated medication guidelines.

[0100] S220. Based on the sebum secretion gradient characteristics of each facial grid, a matching analysis is performed on the medication timestamp, application area, and acne medication application amount to obtain the sebum matching medication execution index, which includes sebum matching time compliance, sebum matching spatial overlap, and sebum matching dosage compliance.

[0101] Based on the acne drug identifier in the facial medication execution data, the drug barcode is scanned and matched against a pre-set acne topical drug knowledge base to obtain the general usage guidelines for the acne drug, including recommended usage time, applicable skin type, and standard dosage range per unit area.

[0102] The Acne Topical Drug Knowledge Base is a standardized database of topical acne drugs built upon the "Chinese Guidelines for the Treatment of Acne," official instructions for topical acne drugs filed with the National Medical Products Administration, and consensus statements on clinical drug use in dermatology. Information within the knowledge base is categorized and archived according to drug pharmacological classification, with each entry linked to an acne drug identifier. The core fields are determined according to the following rules: Recommended usage period: Based on drug photostability and skin absorption and metabolism patterns, the optimal clinical usage window is extracted from guidelines and drug instructions; Applicable skin type: Based on the drug's mechanism of action and irritation characteristics, combined with guidelines, the appropriate facial sebum secretion level is determined; Standard dosage range per unit area: Based on the standard single-use dosage recommended in the drug instructions for the entire face, the dosage per unit area is calculated using the average adult facial area, and the standardized range is determined by matching the fixed area of ​​the single-face grid in this protocol.

[0103] This embodiment takes the commonly used 0.1% retinoic acid cream as an example. The logic for determining the corresponding information is as follows: First, the recommended usage time: based on the photosensitivity of retinoic acid drugs, both guidelines and drug instructions clearly recommend using it at night in the dark. Therefore, the recommended usage time is determined to be 20:00-22:00 daily. Second, the applicable skin type: based on the core mechanism of the drug in controlling oil and regulating follicular keratinization, guidelines recommend its use for oily and combination acne-prone skin. Third, the standard dosage range per unit area: based on the standard dosage of 0.5g recommended in the drug instructions for a single full-face application, combined with the average adult facial area of ​​approximately 400cm², the standard dosage range per unit area suitable for a single facial grid is finally determined to be 0.1mg / cm²-0.5mg / cm². This ensures that the extracted drug's general usage specifications are accurate and compliant, and completely consistent with clinical drug usage specifications, and can be directly used for dynamic adjustment of the subsequent oil matching execution benchmark.

[0104] Based on the high and low sebum secretion gradient characteristics of each facial grid determined by S210, the general drug use guidelines are dynamically adjusted to generate a corresponding sebum matching execution benchmark for each facial grid. This sebum matching execution benchmark is used to characterize the optimal execution standard for topical acne drugs under different facial sebum secretion states, including the time execution benchmark, spatial coverage benchmark, and dosage execution benchmark that match the sebum secretion gradient of the facial grid.

[0105] Based on a knowledge base of topical acne medications, matching drug photostable and skin absorption and metabolism characteristics, time-based application benchmarks are established for different sebum secretion gradients. For facial grids with high sebum secretion gradients, the time-based application benchmark is the core evening application period of the recommended usage time, denoted as... ,in For the start time of the period, This refers to the end time of the time period (this example uses 0.1% retinoic acid cream as an example). =[20:00 22:00]), is the only compliant medication time for facial meshes with high sebum secretion gradients;

[0106] For facial meshes with low sebum secretion gradients: The time of application is based on the recommended morning time slot. =[ [Usage during evening hours] =[ Both time periods are compliant medication use periods.

[0107] Let the timestamp of a single medication event be . The compliance time window is [duration]. = (Unit: hours), the minimum absolute deviation between the actual medication time and the time execution benchmark is [duration]. The compliance of oil matching time for a single facial grid. The calculation formula is:

[0108] ;

[0109] in, The oil matching time compliance of the i-th facial grid is defined by a value in the range of [0,1]. It is used to quantitatively characterize the degree of matching between the actual medication time and the personalized time execution benchmark corresponding to the facial grid. The closer the value is to 1, the higher the degree of matching between the medication time and the standard requirements and the better the time compliance. The timestamp of a single medication event falls entirely within the time execution benchmark corresponding to the facial grid. In this case, the medication time fully complies with the requirements of the oil matching execution benchmark, and the oil matching time compliance score is 1.

[0110] Additionally, for gradient grids targeting low sebum secretion... Pick and , The minimum value of the deviation duration, ultimately Take the maximum value of the oil matching time compliance calculation results from the two time periods.

[0111] The oil matching time compliance calculation function implements a gradient mapping between deviation duration and compliance score: a perfect score of 1 is given if the deviation falls completely within the compliance window; the score decreases linearly within a window of 1; the score decreases rapidly between a deviation of 1 and 3 window lengths; and the score is 0 if the deviation exceeds 3 window lengths. This function can accurately distinguish compliance differences of different deviation levels (e.g., within the 20:00-22:00 compliance window, the compliance score for medication use at 19:00 is 0.5, and the compliance score for medication use at 18:00 is 0.25).

[0112] For a single facial mesh, a spatial coverage benchmark based on the sebum secretion gradient characteristics is first determined. This spatial coverage benchmark is used to determine whether the facial mesh should be included in the application area and has a direct mapping relationship with the results of the lesion aggregation area division.

[0113] For facial grids with high sebum secretion gradients, the spatial coverage benchmark is regional barrier coverage. The determination rule is as follows: if the facial grid is covered by a lesion cluster area, or if the total number of acne lesions within the grid is ≥2, then the facial grid must be included in the application area. Simultaneously, other facial grids with high sebum secretion gradients that are spatially adjacent to the aforementioned core lesion coverage grid (eight neighboring grids, referring to eight standardized facial grids directly adjacent vertically, horizontally, and diagonally to the core lesion coverage grid, completely corresponding to the square standardized facial grid, ensuring continuous coverage without gaps in physical space) must also be included in the application area as auxiliary barrier coverage grids, regardless of whether they themselves contain acne lesions. High-sebum grids without visible acne lesions, not covered by lesion cluster areas, and not adjacent to any core lesion coverage grid are not subject to mandatory inclusion. This setting is based on the fact that in a high sebum secretion state, sebum flushing is significant. If only isolated lesion grids are covered, the drug is easily diluted and cleared by the surrounding active sebum. It is necessary to form a continuous drug barrier by covering adjacent grids to maintain the effective drug concentration in the core lesion area.

[0114] For facial meshes with low sebum secretion gradients, the spatial coverage benchmark is the coverage of the lesion mesh. The determination rule is: if the facial mesh is covered by an area with concentrated lesions, or if there are acne lesions within the facial mesh, then the facial mesh must be included in the application area; low-sebum meshes that are not covered and have no visible acne lesions are not required to be included. This setting is based on the fact that the skin barrier is relatively fragile in a low sebum secretion state, and only meshes containing lesions need to be precisely covered to avoid over-application and irritation to healthy skin, thus achieving the dual goals of treatment and protection.

[0115] Let X be the application selection status of the acne patient on this facial mesh (X=1 means the mesh is selected for application, X=0 means it is not selected), and Y be the spatial coverage baseline attribute of the facial mesh (Y=1 means the facial mesh must be included in the application area, Y=0 means the facial mesh is not required to be included in the application area). Let Y be the oil matching spatial overlap of the facial mesh. The formula for calculating (single-grid smear compliance) is:

[0116] ;

[0117] When Y=1, corresponding to the high sebum secretion gradient grid or the low sebum lesion core grid (mandatory coverage area), the sebum matching spatial overlap is directly determined by the actual application status X. Applying as required earns full marks, while missing application earns 0 marks. When Y=0, corresponding to the low sebum normal non-lesion grid (unnecessary non-coverage area), not applying earns full marks, while applying without authorization deducts 0.5 marks, reflecting a safety orientation that discourages excessive application. The value range is [0,1], which is used to quantitatively characterize the spatial matching degree between the actual application area of ​​the facial mesh and the personalized specification requirements for oil matching.

[0118] For facial meshes with high sebum secretion gradients, the dosage reference is the upper limit of the standard dosage range per unit area; for facial meshes with low sebum secretion gradients, the dosage reference is the lower limit of the standard dosage range per unit area. Let the dosage of acne medication for a single mesh be... The relative deviation ratio between the amount of acne medication applied and the standard application amount is: Compliance of oil matching dosage in a single grid Calculated according to the following rules:

[0119] ;

[0120] Among them, when hour, 1. The actual dosage of acne medication applied is within ±20% of the standard dosage, which is within the clinically acceptable error range. Therefore, the dosage is considered fully compliant, and the compliance score for oil-matching dosage is set at 1. When 0.2 < When ≤1, =1− When the deviation exceeds the acceptable range but does not exceed 100%, the compliance of the oil matching dosage decreases linearly with the deviation ratio, accurately distinguishing the compliance differences of different deviation levels; when When >1, =0: Deviation exceeding 100% (actual acne medication application amount is less than 50% of the baseline dosage or exceeds twice the baseline dosage) is judged as serious non-compliance, and the compliance score for oil matching dosage is 0; serious underdosing cannot achieve the effective therapeutic concentration, while excessive dosage significantly increases the risk of skin irritation, both of which do not comply with the principles of safe medication use.

[0121] The threshold of 0.2 is in line with the "Chinese Guidelines for Acne Treatment" and the clinical use guidelines for topical dermatological drugs. Clinical studies of topical acne drugs (retinoic acid, benzoyl peroxide, topical antibiotics, etc.) have confirmed that a dosage deviation of ±20% will not affect the effective drug concentration in the pilosebaceous unit, nor will it increase the risk of adverse reactions such as skin erythema, desquamation, and stinging. This is an acceptable range of medication error recognized by dermatologists both domestically and internationally.

[0122] S230. Based on the dynamic trend of acne lesion type and number in acne patients, a modified disease weight is assigned to each type of acne lesion according to the difference in clinical severity.

[0123] The location coordinates of each acne lesion in the current period are extracted from facial acne feature data. Based on spatial proximity, the lesions are divided into several lesion clusters. Each lesion cluster corresponds to a continuous dense area of ​​lesions on the face. In this embodiment, the spatial proximity is determined by the rule that two acne lesions with an Euclidean distance of less than 3 mm are considered to belong to the same lesion cluster. This 3 mm threshold matches the clinical tendency of acne lesions to merge. That is, when the distance between acne lesions is less than 3 mm, the risk of inflammatory expansion and fusion of acne lesions to form a larger lesion area increases significantly, which meets the dermatological clinical definition standard for the risk of acne clustering.

[0124] For each lesion cluster area, the number of acne lesions of each type within the cluster area is counted. Static baseline disease weights are preset for each type of acne lesion according to clinical severity, with higher severity resulting in greater weights. The total static baseline disease weight for that lesion cluster area is then calculated. In this embodiment, according to the clinical grading standards of the Chinese Acne Treatment Guidelines, acne lesions are divided into three categories: non-inflammatory comedones, inflammatory papules and pustules, and nodules and cysts. The preset static baseline disease weights are as follows: non-inflammatory comedones have a weight of 1, inflammatory papules and pustules have a weight of 3, and nodules and cysts have a weight of 5. These weights are assigned based on the clinical differences in the degree of damage to the pilosebaceous unit, the intensity of the inflammatory response, and the risk of scarring after healing for each type of lesion.

[0125] The location coordinates of each acne lesion in the previous period are extracted from facial acne feature data. The lesion cluster areas in the current period are spatially matched with those in the previous period. The total number of acne lesions in the same cluster area in adjacent periods is compared to determine the trend of disease change: an increase in the number of lesions (the total number of acne lesions in the cluster area in the current period increases by more than 20% compared to the previous period) indicates disease progression; a decrease in the number of lesions (the total number of acne lesions in the cluster area in the current period decreases by more than 20% compared to the previous period) indicates lesion regression; and no change in the number of lesions (the total number of acne lesions in the cluster area in the current period changes by less than ±20% compared to the previous period) indicates disease stability. A correction coefficient for the number of lesions is assigned to different trends: 1.3 for disease progression, 0.8 for lesion regression, and 1 for disease stability. The correction coefficient for the number of lesions is set based on the dynamic evolution of clinical acne. When the disease progresses, the difficulty of drug penetration and absorption increases, so the weight needs to be increased to reflect the increased difficulty of treatment. During the lesion regression period, the lesions tend to heal, and the need for drug action decreases, so the weight can be appropriately decreased.

[0126] The static baseline disease weight of each skin lesion cluster area is multiplied by the corresponding quantity change correction coefficient to obtain the corrected disease weight of that skin lesion cluster area, which is used to characterize the overall disease severity of that skin lesion cluster area in the current cycle.

[0127] S240. Based on the spatial distribution of different types of acne lesions on the face of acne patients, calculate the degree of lesion clustering in the region and generate the regional clustering coefficient; integrate the corrected disease weight with the regional clustering coefficient to generate the grid acne disease index.

[0128] The density of acne lesions within the current period's lesion cluster area is calculated to generate a regional clustering coefficient. This coefficient characterizes the concentration risk of lesions within the cluster area; the denser the acne lesions, the higher the regional clustering coefficient. In this embodiment, the regional clustering coefficient is calculated based on the number of acne lesions per unit area within the cluster area, using the following formula: ,in This is the regional clustering coefficient. This refers to the total number of all acne lesions within the area where the lesions cluster. The area of ​​the smallest circumscribed convex polygon of the lesion cluster region is expressed in cm². As a clinical reference for density threshold, this embodiment sets it to 2 lesions / cm² (i.e., when there are 2 acne lesions per square centimeter, it is considered a density risk benchmark) based on the "Chinese Guidelines for Acne Treatment" and dermatological clinical consensus; the clustering coefficient of this area These are dimensionless values, reflecting the ratio of actual skin lesion density to the clinical reference baseline. A value greater than 1 indicates that the density exceeds the benchmark and the risk of clustering is high. A value less than 1 indicates that the density is below the benchmark and the risk of clustering is low.

[0129] The corrected disease weight of each lesion cluster area is multiplied by the corresponding regional clustering coefficient to obtain the comprehensive disease risk value of that lesion cluster area. The comprehensive disease risk value of each lesion cluster area is then assigned to the corresponding facial grid according to the spatial coverage relationship between each facial grid and the lesion cluster area, generating a grid acne disease index for each facial grid. This index is used to characterize the severity of the disease and the clustering risk of the lesion cluster area where each facial grid is located. If a facial grid is covered by multiple lesion cluster areas, the maximum comprehensive disease risk value of each covered lesion cluster area is taken as the grid acne disease index of that facial grid, reflecting the principle of maximizing risk.

[0130] Behavioral assessment module: Based on facial oil feature data, it integrates and analyzes the grid acne condition index and the oil-matching medication execution index to output the facial grid-based drug utilization rate results.

[0131] In this embodiment, as Figure 3 As shown, based on facial oil feature data, a fusion analysis is performed on the acne condition index and the oil-matching medication execution index to output the facial grid-based drug utilization rate results, including the following specific steps:

[0132] S310. Based on the sebum secretion gradient characteristics of each facial grid in acne patients, the sebum-matched medication execution index is weighted and fused, and dynamically calibrated in combination with the grid acne condition index to calculate the drug utilization score of each facial grid.

[0133] In this embodiment, the sebum secretion gradient feature is divided into two levels: high sebum secretion gradient and low sebum secretion gradient, corresponding to two sets of weight allocation schemes: for facial meshes with high sebum secretion gradient, the weight for sebum matching time compliance is 0.2, the weight for sebum matching spatial overlap is 0.3, and the weight for sebum matching amount compliance is 0.5; for facial meshes with low sebum secretion gradient, the weight for sebum matching time compliance is 0.2, the weight for sebum matching spatial overlap is 0.5, and the weight for sebum matching amount compliance is 0.3.

[0134] The weighting scheme is based on the following: In a high sebum secretion state, excessive sebum secretion dilutes the drug and accelerates its clearance. Sufficient dosage is the decisive factor in maintaining the effective drug concentration in the pilosebaceous unit. Therefore, dosage compliance has the highest weight. In a low sebum secretion state, the skin barrier is relatively fragile, and excessive application can easily cause irritation. Precise control of the application area and covering only the lesion area is the key to ensuring clinical benefits and safety. Therefore, spatial overlap has the highest weight. Temporal compliance reflects the photostability and absorption rhythm of the drug and has a weak correlation with sebum state. Therefore, it maintains a fixed base weight of 0.2 in different sebum gradients.

[0135] Based on the weight allocation scheme corresponding to the sebum secretion gradient features of the facial mesh, the sebum matching time compliance is determined. Oil-oil matching space overlap and compliance of oil and fat dosage The basic behavioral score of the facial mesh is calculated by weighted summation of the three indicators. The calculation formula is:

[0136] ;

[0137] in, , , These are weights for compliance of oil matching time, overlap of oil matching space, and compliance of oil matching dosage. Specific values ​​are selected based on the oil secretion gradient of the facial mesh, along with the basic behavioral score. The value range is [0,1], reflecting the theoretical contribution of the quality of drug administration to drug utilization under this lipid secretion state;

[0138] Based on the facial grid acne severity index The disease penetration loss coefficient was calculated. The baseline behavioral scores are dynamically calibrated to calculate the drug utilization score; in this embodiment, the disease penetration reduction coefficient is used. The calculation formula is: ,in The condition index for reticular acne is dimensionless. The disease severity factor is set to 0.2 in this embodiment. This value is based on quantitative research in clinical pharmacology on the difficulty of drug penetration into acne lesions. When the reticuloacne disease index... When =1 (corresponding to the baseline condition of a single non-inflammatory pimple), the drug penetration loss is approximately 18%. ≈0.82); when When the value is 5 (corresponding to severe lesions with clusters of inflammatory papules and pustules), the drug penetration loss is approximately 63%. ≈0.37), meaning the drug utilization rate decreased to less than 40%; this decay curve conforms to the nonlinear characteristics observed clinically, where the loss is gradual in mild cases and rapidly accelerates in severe cases.

[0139] The overall setting of the exponential decay function is based on the following: as the severity of the disease increases, the difficulty of drug penetration and bioavailability in lesions with inflammatory infiltration, hyperkeratosis, and hair follicle blockage increases non-linearly. The more severe the disease, the more significant the degradation, and the degradation effect accelerates as the disease worsens. A value of 0.2 places the loss factor within the common range of the disease index. The system can reasonably differentiate drug availability levels at different severity levels.

[0140] The drug utilization score of the facial grid is obtained by multiplying the basic behavioral score by the disease penetration reduction coefficient. The value range is [0,1]. This score is used to quantitatively characterize the expected clinically effective drug utilization level determined by the actual drug use behavior execution quality of the facial grid under the current sebum secretion state and disease background.

[0141] S320. Classify drug utilization rate levels based on the drug utilization rate scores of each facial grid, and identify the core influencing factors that lead to insufficient drug utilization rate of the facial grid based on the decomposition results of disease impairment loss and behavioral deviation loss.

[0142] The drug utilization rate score of each facial grid is compared with a preset scoring threshold to classify it into two levels: high and low. In this embodiment, the preset scoring threshold is set to 0.6. Facial grids with a drug utilization rate score ≥ 0.6 are classified as having a high drug utilization rate, while those with a score < 0.6 are classified as having a low drug utilization rate. This threshold is set based on the fact that the drug utilization rate is determined by both the quality of medication administration and the degree to which the patient's condition restricts drug penetration. When the score is below 0.6, if there is no impairment of the patient's condition (patient penetration impairment coefficient)... =1), which means the basic behavioral score is 1. A score below 0.6 indicates a significant defect in medication use; if the quality of medication use is perfect, it means that the disease loss coefficient is below 0.6, indicating that the severity of the lesions has led to a drug penetration loss of more than 40%. Both of these situations suggest that the drug utilization level of the facial mesh has dropped to the point where targeted intervention is required. Therefore, 0.6 is used as the threshold for initiating intervention analysis.

[0143] For target facial meshes with low drug utilization levels, calculate the total loss value of drug utilization score. The total loss value is the sum of the full score and the drug utilization score. The difference is calculated using the following formula: To accurately identify the core sources of drug utilization loss, the total loss value was broken down into independent disease-related impairment losses. Loss of behavioral deviation The study consists of two parts: disease severity loss, which reflects the objective limiting effect of lesion severity on drug penetration and utilization; and behavioral deviation loss, which reflects the utilization loss caused by patients' deviation from personalized medication norms.

[0144] Among them, the loss due to illness The calculation formula is: Basic Behavioral Rating (BBR) Multiply by the disease penetration loss coefficient The corresponding loss ratio, behavioral deviation loss The calculation formula is: = That is, the difference between the baseline behavioral score and the full score, and the sum of the two losses satisfies ;

[0145] The preset threshold for determining the dominant loss type is 0.5, and the loss is determined based on the severity of the illness. Loss of behavioral deviation The proportion of the two types of losses in the total loss value determines the dominant loss type of the target facial mesh: the proportion of loss due to disease damage. The loss was determined to be primarily caused by the illness; the rest were determined to be primarily caused by behavior.

[0146] For target facial meshes with different loss-dominant types, the corresponding core influencing factor identification rules are matched: for behavior-dominant meshes, the single-dimensional contribution of three medication compliance indicators to behavioral deviation loss is calculated, and the dimension with the largest contribution is determined as the core influencing factor;

[0147] The contribution of a single dimension is measured by the deviation of the indicator, which is the difference between each compliance indicator and the full score of 1. The calculation formula is as follows: ,Pick , , The dimension corresponding to the maximum value in the data is used as the core influencing factor, which corresponds to the time deviation type, spatial deviation type, and dosage deviation type, respectively. For the disease-dominant grid, the severity of the lesion and the difficulty of drug penetration are identified as the core influencing factors, specifically labeled as the disease-constrained type. The drug utilization rate level, loss-dominant type, and core influencing factors of each facial grid are uniformly used as the output results of the behavior assessment module and transmitted to the subsequent intervention generation module.

[0148] Intervention generation module: Based on the results of facial grid-based drug utilization, it generates medication execution planning instructions that match the individual facial region characteristics of acne patients.

[0149] In this embodiment, as Figure 4 As shown, based on the facial mesh-based drug utilization results, medication execution planning instructions are generated to match the individual facial region characteristics of acne patients, including the following specific steps:

[0150] S410. Aggregate spatially adjacent facial grids with consistent core influencing factors to form clinically interventionable facial region units.

[0151] The pre-defined rule for determining the spatial adjacency of the standardized facial mesh is the 4-neighbor rule, which states that for any target facial mesh, only the meshes directly adjacent in the four directions of top, bottom, left, and right are considered spatially adjacent, excluding diagonal adjacency. This rule matches the continuous anatomical structure of the facial skin, ensuring that the aggregated regional units are physically continuous and unbroken, facilitating the accurate mapping and execution of subsequent clinical intervention instructions.

[0152] Using the core impact factor as the core aggregation condition, and combined with the drug utilization rate level, target facial meshes that simultaneously meet the conditions of spatial adjacency, belong to the low drug utilization rate level, and have the same core impact factor are clustered and merged.

[0153] The core influencing factors include four categories: time deviation, space deviation, dosage deviation, and disease-related factors. Drug utilization rate is divided into two levels: low utilization rate and high utilization rate. The aggregation operation is only performed on the target facial mesh of the low utilization rate level. The mesh of the high utilization rate level does not participate in the aggregation, so no targeted intervention is required.

[0154] The specific aggregation process uses a connected component labeling algorithm. First, face grids with the same core influence factor type are selected from all low-utilization level grids as seed points. Then, according to the 4-neighborhood rule, the grids are expanded outward layer by layer to include all low-utilization level grids that are spatially connected to the seed points and have the same core influence factor type into the same cluster until there are no more adjacent grids to expand.

[0155] For isolated target facial grids with low drug utilization levels that cannot be merged, i.e., there are no other low-utilization-level grids around the grid or the core influencing factors of adjacent grids are inconsistent, they are assigned to adjacent intervention regions of the same type according to the core influencing factors. If there are multiple intervention regions of the same type with the same core influencing factors in adjacent regions, the one with the closest spatial distance is selected for merging, ensuring that all low-utilization grids on the face are covered and the intervention regions are divided continuously and completely. Finally, several continuous and clinically interventionable facial intervention region units are formed. Each facial intervention region unit consists of multiple spatially continuous low-utilization standardized facial grids, with a unified core influencing factor, a unified drug utilization hierarchy, and a unique clinical intervention target.

[0156] S420. Based on the drug utilization rate level and core influencing factors of each facial region unit, match the corresponding drug administration execution plan instruction for each facial region unit.

[0157] Based on the core influencing factor type of each facial intervention area unit, the corresponding basic intervention strategy is matched. The core influencing factors include four types: time deviation type, spatial deviation type, dosage deviation type, and disease-constrained type, which correspond to different directions of basic intervention strategies: time deviation type is matched with medication timing calibration strategy, spatial deviation type is matched with application range guidance strategy, dosage deviation type is matched with medication dosage control strategy, and disease-constrained type is matched with disease penetration assistance strategy.

[0158] The matched basic intervention strategy, combined with the facial anatomical coverage and drug utilization level of the facial intervention area unit, is transformed into a medication execution plan instruction that patients can intuitively understand.

[0159] The facial anatomical coverage is determined based on the set of standardized facial grid coordinates contained in the facial intervention area unit. This is then converted into a natural language description of the facial region. The drug utilization rate level indicates that the facial intervention area unit belongs to the low utilization rate level and requires special attention. The medication execution plan instructions for each facial region are summarized, and basic nursing and follow-up requirements are supplemented based on the patient's overall facial condition to form the overall personalized medication execution plan instructions for the current cycle unit.

[0160] Embodiments of the present invention also provide an electronic device, including a memory, a processor, and a communication bus; the memory and the processor are connected via the communication bus. The memory stores an acne patient information planning system based on the IMB model, which can be loaded and executed by the processor, as provided in the above embodiments.

[0161] The memory can be used to store instructions, programs, code, code sets, or instruction sets. The memory may include a program storage area and a data storage area. The program storage area may store instructions for implementing an operating system, instructions for at least one function, and instructions for implementing the IMB-based acne patient information planning system provided in the above embodiments. The data storage area may store data involved in the IMB-based acne patient information planning system provided in the above embodiments.

[0162] A processor may include one or more processing cores. The processor executes instructions, programs, code sets, or instruction sets stored in memory, and calls data stored in memory to perform various functions and process data as described in this application. The processor may be at least one of a specific application-specific integrated circuit, a digital signal processor, a digital signal processing device, a programmable logic device, a field-programmable gate array, a central processing unit, a controller, a microcontroller, and a microprocessor. It is understood that, for different devices, the electronic devices used to implement the above-described processor functions may also be other types, and the embodiments of this application do not specifically limit the specific implementation.

[0163] A communication bus may include a pathway for transmitting information between the aforementioned components. The communication bus can be a PCI bus or an EISA bus, etc. Communication buses can be categorized into address buses, data buses, control buses, etc.

[0164] This application provides a computer-readable storage medium storing a computer program that can be loaded by a processor and executed as described in the above embodiments, which is an acne patient information planning system based on the IMB model.

[0165] In this embodiment, a computer-readable storage medium can be a tangible device that holds and stores instructions used by an instruction execution device. A computer-readable storage medium can be, but is not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any combination thereof. Specifically, a computer-readable storage medium can be a portable computer disk, a hard disk, a USB flash drive, a random access memory, a read-only memory, an erasable programmable read-only memory, a podium random access memory, a portable compressed disk read-only memory, a digital multifunction disk, a memory stick, a floppy disk, an optical disk, a magnetic disk, a mechanical encoding device, or any combination thereof.

[0166] The terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0167] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the foregoing application concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions claimed in this application.

Claims

1. An information planning system for acne patients based on the IMB model, characterized in that, Includes the following modules: Data acquisition module: Simultaneously collects facial acne characteristic data, facial medication administration data, and facial oil characteristic data of acne patients; Feature extraction module: Based on facial oil feature data and facial medication execution data, generate oil-matching medication execution index; based on acne lesion type and spatial clustering features in facial acne feature data, generate grid acne condition index. Behavioral assessment module: Based on facial oil feature data, it integrates and analyzes the grid acne disease index and the oil-matching medication execution index to output the facial grid-based drug utilization rate results; Intervention generation module: Based on the results of facial grid-based drug utilization, it generates medication execution planning instructions that match the individual facial region characteristics of acne patients.

2. The acne patient information planning system based on the IMB model according to claim 1, characterized in that, Based on facial oil feature data and facial medication execution data, an oil-matching medication execution index is generated, including the following steps: Based on the facial oil feature data of acne patients, the oil secretion gradient features corresponding to each facial grid of acne patients are obtained; Based on the sebum secretion gradient characteristics of each facial grid, a matching analysis was performed on the medication timestamp, application area, and amount of acne medication applied to obtain the sebum matching medication execution index, which includes the sebum matching time compliance, sebum matching spatial overlap, and sebum matching dosage compliance.

3. The acne patient information planning system based on the IMB model according to claim 2, characterized in that, Matching analyses were performed on medication timestamps, application areas, and acne medication dosages to obtain an oil-matching medication execution index that includes compliance with oil-matching time, spatial overlap, and dosage. This process includes the following steps: Based on the sebum secretion gradient characteristics of each facial grid, the sebum matching execution benchmark corresponding to that facial grid is determined; the sebum matching execution benchmark includes the time execution benchmark, the spatial coverage benchmark, and the dosage execution benchmark. The medication timestamps of acne patients are matched and analyzed with the time execution benchmark of the corresponding facial grid to obtain the oil matching time compliance of the facial grid. The coverage compliance of the application area of ​​each medication event of acne patients was judged based on the spatial coverage benchmark, and the oil matching spatial overlap of the facial grid was obtained. By matching the dosage of acne medication applied by acne patients with the dosage implementation benchmark, the compliance of the oil matching dosage for this facial grid was obtained.

4. The acne patient information planning system based on the IMB model according to claim 3, characterized in that, Based on the acne lesion types and spatial clustering characteristics in facial acne feature data, a grid acne condition index is generated, including the following steps: Based on the dynamic trends of acne lesion types and numbers in acne patients, adjusted disease weights were assigned to each type of acne lesion according to differences in clinical severity. Based on the spatial distribution of different types of acne lesions on the face of acne patients, the degree of lesion clustering in the region is calculated to generate a regional clustering coefficient. The corrected disease weights are then fused with the regional clustering coefficients to generate a grid acne disease index.

5. The acne patient information planning system based on the IMB model according to claim 4, characterized in that, Simultaneously collect facial acne characteristic data, facial medication administration data, and facial oil characteristic data from acne patients, including the following steps: Facial image data of acne patients were collected using image recognition algorithms to construct facial acne feature data, which includes acne lesion type, acne lesion count, and acne lesion location coordinates. Based on medication events of acne patients, facial medication execution data is generated, including medication timestamps, application areas, acne drug identification, and acne drug application dosage. We collected facial sebum secretion rate data from acne patients and constructed facial sebum feature data to characterize the sebum secretion characteristics of each facial grid.

6. The acne patient information planning system based on the IMB model according to claim 5, characterized in that, Based on facial oil characteristics data, a fusion analysis was performed on the acne condition index and the oil-matching medication execution index to output the facial grid-based drug utilization rate results, including the following steps: Based on the sebum secretion gradient characteristics of each facial grid in acne patients, the sebum-matched medication execution index is weighted and fused, and dynamically calibrated in combination with the grid acne condition index to calculate the drug utilization score of each facial grid. The drug utilization rate was classified into levels based on the drug utilization rate score of each facial grid, and the core influencing factors leading to insufficient drug utilization of the facial grid were identified based on the decomposition results of disease impairment loss and behavioral deviation loss.

7. The acne patient information planning system based on the IMB model according to claim 6, characterized in that, Dynamic calibration is performed using the reticular acne condition index, including the following steps: Based on the limiting effect of acne lesion severity, as represented by the grid acne severity index of each facial grid, on drug penetration and utilization, the baseline behavioral score obtained by weighted fusion of the oil-matched drug administration execution index is negatively calibrated to obtain the drug utilization score of the facial grid.

8. The acne patient information planning system based on the IMB model according to claim 7, characterized in that, Based on the facial mesh-based drug utilization results, a medication execution plan is generated that matches the individual facial region characteristics of acne patients, including the following steps: Adjacent facial grids with consistent core influencing factors are aggregated to form clinically interventionable facial region units; Based on the drug utilization rate level and core influencing factors of each facial region unit, a corresponding drug administration plan instruction is matched for each facial region unit.

9. 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 implements the acne patient information planning system based on the IMB model as described in any one of claims 1-8 by calling the computer program stored in the memory.

Images