AI-based skin analysis and personalized prescription recommendation method and system

Abstract

In the AI-based skin analysis and customized prescription recommendation method of the present invention, the method comprises the steps of: a specialty server transmitting image data containing customer data to a service provider server to request analysis data; a service provider server generating analysis data based on the image data and transmitting it to the specialty server; a specialty server requesting the service provider server to generate customized prescription data corresponding to the image data and analysis data; and a service provider server generating customized prescription data and transmitting it to the specialty server. The step of the service provider server generating analysis data based on the image data and transmitting it to the specialty server comprises the steps of: the service provider server performing preprocessing of the image data; the service provider server inputting the preprocessed image data into an artificial intelligence model capable of classifying multiple symptoms to extract major symptom data and secondary symptom data; the service provider server generating comprehensive skin condition data based on the symptom data; and the service provider server inputting the comprehensive skin condition data and the customer data into a time-series prediction model to generate analysis data including information on symptom changes and improvements over a future period. It includes an additional step of transmitting to the specialty store server.

Description

Technology Field

[0001] The present invention relates to an AI-based skin analysis and customized prescription recommendation method and system that inputs skin condition and captured image data, performs AI-based analysis on a server, and automatically recommends an optimized customized prescription based on time series prediction and clinical data. Background Technology

[0003] The diagnosis of skin diseases or cosmetic skin conditions has generally been performed by skilled professionals through visual evaluation, questionnaires, gross inspection, and quantitative analysis using specific medical devices. However, these existing diagnostic methods often rely on the subjective judgment of experts and have limitations in that it is difficult to make a comprehensive assessment when multiple symptoms are present simultaneously. Furthermore, even for the same patient, diagnostic results can vary depending on environmental factors such as lighting, equipment, and shooting angles, which can lead to issues with diagnostic precision and consistency. Previously, it was common practice to measure skin conditions by linking software optimized for specific hardware, such as skin scanners and diagnostic sensors, and for experts to provide customized prescriptions based on manual methods. However, this approach is dependent on specific devices, has limitations in the automation and personalization of prescriptions, and struggles to incorporate data-driven predictive functions, resulting in difficulties in continuous condition tracking and predicting treatment effectiveness.

[0004] Recently, with the advancement of artificial intelligence technology, particularly deep learning-based image analysis and time-series forecasting, there is a growing demand for technologies that quantitatively analyze image data collected from users, predict treatment progress, and automatically generate personalized prescriptions. However, existing technologies still exhibit limitations, such as reliance on sensors or devices, manual prescription generation, and analysis structures that fail to reflect the interactions between multiple symptoms. Consequently, there is an increasing need for an integrated and automated cloud-based system that standardizes skin images captured by various devices for device-independent analysis, and encompasses quantified symptom classification, time-series forecasting, and clinical data-based prescription recommendations.

[0005] Prior art includes Korean Registered Patent No. 10-2365783 (Method and apparatus for generating AI-based customized prescription data), but it merely includes the steps of receiving a user's skin image, receiving the user's skin measurement data, inputting the skin image into a first model based on an artificial neural network to calculate a first degree of membership for a skin type corresponding to the user, inputting the skin measurement data into a second model based on an artificial neural network to calculate a second degree of membership for a skin type corresponding to the skin measurement data, determining a final degree of membership based on the first degree of membership and the second degree of membership, and determining customized prescription data corresponding to the user based on the final degree of membership. The problem to be solved

[0007] To address the aforementioned problems, the objective of the present invention is to provide an AI-based skin analysis and personalized prescription recommendation method and system capable of simultaneously resolving the issues of existing subjective diagnosis methods, hardware dependency, limitations of manual prescription generation, and the lack of prediction-based treatment planning by performing device-independent analysis through the standardization of skin images input from various image acquisition devices, predicting the future progression of the skin condition based on quantitative symptom classification and time-series prediction, and automatically generating an optimal personalized prescription utilizing clinical data. means of solving the problem

[0009] In the AI-based skin analysis and customized prescription recommendation method of the present invention, the method comprises the steps of: a specialty server transmitting image data containing customer data to a service provider server to request analysis data; a service provider server generating analysis data based on the image data and transmitting it to the specialty server; a specialty server requesting the service provider server to generate customized prescription data corresponding to the image data and analysis data; and a service provider server generating customized prescription data and transmitting it to the specialty server. The step of the service provider server generating analysis data based on the image data and transmitting it to the specialty server comprises the steps of: the service provider server performing preprocessing of the image data; the service provider server inputting the preprocessed image data into an artificial intelligence model capable of classifying multiple symptoms to extract major symptom data and secondary symptom data; the service provider server generating comprehensive skin condition data based on the symptom data; and the service provider server inputting the comprehensive skin condition data and the customer data into a time-series prediction model to generate analysis data including information on symptom changes and improvements over a future period. It includes an additional step of transmitting to the specialty store server.

[0010] The AI-based skin analysis and customized prescription recommendation system of the present invention comprises: a specialty store server that requests analysis data by transmitting image data containing customer data to a service provider server and requests the generation of customized prescription data corresponding to the image data and analysis data; and a service provider server that generates analysis data based on the image data and transmits it to the specialty store server, and generates customized prescription data and transmits it to the specialty store server. The service provider server performs preprocessing of the image data, inputs the preprocessed image data into an artificial intelligence model capable of classifying multiple symptoms to extract major symptom data and secondary symptom data, generates comprehensive skin condition data based on the symptom data, inputs the comprehensive skin condition data into a time series prediction model to generate analysis data including information predicting symptom changes and improvements over a certain period in the future, and transmits it to the specialty store server. Effects of the invention

[0012] According to one embodiment of the present invention, consistent analysis of skin images is possible regardless of the device, so analysis services of the same quality can be provided at specialized stores without the need for separate dedicated equipment.

[0013] In addition, by combining deep learning-based symptom classification and time series prediction models, it is possible to not only quantitatively diagnose the current state but also predict future symptom changes, thereby supporting the establishment of user-customized mid-to-long-term treatment plans.

[0014] Furthermore, by linking clinical big data with customer profile information to automatically recommend multiple optimal prescriptions, it is possible to provide scientifically customized prescriptions that comprehensively consider efficacy, cost, and risk of side effects without the need for manual intervention by experts.

[0015] Furthermore, the present invention can be utilized as a non-medical diagnostic assistance system capable of providing expert-level analysis and prescription support without involving direct diagnostic or treatment acts under medical law, and can simultaneously improve management efficiency and user satisfaction through the automation of the prescription generation and adjustment process. Brief explanation of the drawing

[0017] FIG. 1 is a flowchart illustrating an AI-based skin analysis and customized prescription recommendation method according to an embodiment of the present invention. FIG. 2 is a flowchart illustrating a method for generating analysis data according to an embodiment of the present invention. FIG. 3 is a flowchart illustrating a method for generating customized prescription data according to an embodiment of the present invention. Figure 4 is a configuration diagram of an AI-based skin analysis and customized prescription recommendation system according to an embodiment of the present invention. Specific details for implementing the invention

[0018] Embodiments according to the concept of the present invention may be subject to various modifications and may take various forms; therefore, embodiments are illustrated in the drawings and described in detail in this specification. However, this is not intended to limit the embodiments according to the concept of the present invention to specific disclosed forms, and includes all modifications, equivalents, or substitutions that fall within the spirit and scope of the present invention.

[0019] The terms used herein are used merely to describe specific embodiments and are not intended to limit the invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this specification, terms such as "comprising" or "having" are intended to indicate the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described herein, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0020] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings attached to this specification.

[0022] FIG. 1 is a flowchart illustrating an AI-based skin analysis and customized prescription recommendation method according to an embodiment of the present invention.

[0023] Referring to FIG. 1, a specialty store server (100) requests analysis data by transmitting image data containing customer data to a service provider server (200) (S101). The customer data may include at least one of a plurality of variables that can affect at least one skin condition, such as the customer's age, gender, skin type, past treatment history, skin sensitivity, allergy history, menstrual cycle, lifestyle habits, stress level, sleep duration, dietary habits, and cosmetic usage history, but is not necessarily limited thereto. The service provider server (200) generates analysis data based on the received image data and transmits it to the specialty store server (100) (S103). The specialty store server (100) requests the service provider server (200) to generate customized prescription data corresponding to the image data and analysis data (S105). The service provider server (200) generates customized prescription data and transmits it to the specialty store server (100) (S107).

[0025] FIG. 2 is a flowchart illustrating a method for generating analysis data according to an embodiment of the present invention.

[0026] Referring to FIG. 2, a service providing server (200) performs preprocessing of image data (S201). The service providing server (200) can preprocess image data by performing at least one of normalizing image resolution, color correction, noise removal, and automatic adjustment of brightness and contrast, with the purpose of ensuring consistency between images acquired from various devices and environments. The service providing server (200) can recognize a face or skin area included in the image data and automatically extract a Region of Interest (ROI) that requires analysis.

[0027] A service providing server (200) inputs preprocessed image data into an artificial intelligence model capable of classifying multiple symptoms to extract major symptom data and secondary symptom data (S203). The artificial intelligence model may be composed of at least one of a classification model based on a CNN (Convolutional Neural Network), a deep learning architecture of the ResNet or EfficientNet family, and may be a pre-trained model or a transfer learning model specialized in multi-class classification of skin diseases. The service providing server (200) extracts features from the input image data at the pixel level or patch level, and can determine specific symptoms such as acne, erythema, pigmentation, wrinkles, and loss of elasticity based on the label classification results learned for the corresponding features.

[0028] The service providing server (200) generates comprehensive skin condition data based on the above symptom data (S205). The comprehensive skin condition data may be composed by quantifying or grading various skin indicators, and may include at least one of the following as major components: skin health score, problem area grade, area requiring improvement, symptom density, symptom distribution, complex symptom correlation, and potential symptom risk index, but is not necessarily limited thereto.

[0029] The service providing server (200) inputs the comprehensive skin condition data and the customer data into a time series prediction model to generate analysis data including changes in symptoms and predicted improvement possibilities over a certain period in the future, and transmits it to the specialty store server (100) (S207). By predicting changes in the comprehensive skin condition data at different points along the time axis, the service providing server (200) can quantitatively calculate at least one of the timing when specific skin symptoms will be alleviated or worsened, the fluctuation range of symptoms, and the expected period required for improvement. To this end, customer data including at least one of the customer's age, gender, skin type, stress index, menstrual cycle, and seasonal characteristics can be considered together to construct a time series context that is comparable to a similar group with the same symptoms. The service providing server (200) can perform time series learning and prediction using at least one of a prediction model based on LSTM (Long Short-Term Memory), GRU (Gated Recurrent Unit), or Temporal Convolutional Network, and can also implement an ensemble method that increases prediction reliability by running multiple prediction models in parallel.

[0031] FIG. 3 is a flowchart illustrating a method for generating customized prescription data according to an embodiment of the present invention.

[0032] Referring to FIG. 3, a service providing server (200) analyzes image data and analysis data to extract data on the optimal treatment timing (S301). The service providing server (200) can calculate the optimal timing for the initiation of skin treatment by comprehensively considering the prediction information on major symptom changes included in the analysis data and the biological and environmental factors included in the customer data. The biological factors considered at this time may include at least one of the customer's menstrual cycle, skin regeneration cycle, and skin sensitivity cycle, and the environmental factors may include at least one of the current season, average humidity, UV index, amount of sunlight, and environmental characteristics of the customer's residential area.

[0033] The service providing server (200) generates at least one optimal treatment protocol by matching customer data, optimal treatment timing data, and analysis data included in the image data with past cases of the specialty store clinical data (S303). The service providing server (200) can design a customized treatment protocol that presents an optimal treatment path by comprehensively analyzing at least one of the major symptom types, symptom intensity, predicted symptom change patterns derived from the analysis data, and skin type, age group, gender, medications being taken, and allergy information included in the customer data. The above specialty store clinical data may consist of past performance-based data including at least one of thousands of treatment success / failure cases, symptom-specific improvement rates, side effect incidence rates, and treatment discontinuation rates, and is built in a pre-tagged and normalized form, enabling rapid similarity-based search.

[0034] The service providing server (200) calculates the prescription priority by considering at least one of the expected effect, treatment period, and side effect rate among the optimal treatment protocols, and generates a customized prescription plan corresponding to each protocol (S305). The service providing server (200) can automatically calculate the priority by applying multidimensional evaluation criteria to each of the multiple optimal treatment protocols derived from the protocol generation module (222). The service providing server (200) uses at least one of the expected improvement effect (improvement rate), total treatment period, total cost, expected side effect rate, and customer satisfaction-based score as an evaluation indicator, and the evaluation indicators can be quantified through a weighted integrated scoring process.

[0035] The service providing server (200) generates customized prescription data for each customized prescription plan, including at least one of expected effect, main ingredients, method of use, usage cycle, side effects, and precautions (S307). The service providing server (200) can convert the components of each generated customized prescription plan into a structured data format and process them into prescription data that can be utilized by the specialty store server (100). In one embodiment, the prescription data may include execution guidelines that include at least one of the treatment period for each phase, a list of products to be used within the corresponding period, the application site and dosage of each product, the usage cycle (e.g., daily / three times a week), the time of day (morning / evening), and the method of use (including application and absorption time), and may be configured so that the practitioner or customer of the specialty store can easily understand and follow them. The service providing server (200) transmits at least one customized prescription data generated to the specialty store server (100) (S309).

[0037] Figure 4 is a configuration diagram of an AI-based skin analysis and customized prescription recommendation system according to an embodiment of the present invention.

[0038] Referring to FIG. 4, the AI-based skin analysis and customized prescription recommendation system (10) consists of a specialty store server (100) and a service provision server (200).

[0039] The specialty store server (100) can request analysis data by transmitting image data containing customer data to the service provision server (200) and receive the analysis data. The customer data may include at least one of a plurality of variables that can affect at least one skin condition, such as the customer's age, gender, skin type, past treatment history, skin sensitivity, allergy history, menstrual cycle, lifestyle habits, stress level, sleep duration, dietary habits, and cosmetic usage history, but is not necessarily limited thereto. By transmitting this customer data collected from the customer along with the image data to the service provision server (200), the specialty store server (100) supports more precise and customized analysis. After receiving the analysis data from the service provision server (200), the specialty store server (100) can provide the data in a visualized form to specialty store management personnel, treatment staff, or skin experts through an internal system or user interface. In one embodiment, the analysis data consists of information such as major skin symptoms, secondary symptoms, overall skin condition, and predicted symptom change curves, and can be used to establish a customized treatment schedule or for consultation with a customer.

[0040] Additionally, the specialty store server (100) may request the service provider server (200) to generate customized prescription data. When making a request, it may include not only analysis data but also auxiliary information such as customer preferences (e.g., treatment period, cost range, type of product used) to induce a more realistic prescription plan. This request may be made in real time according to an automated algorithm and may be repeatedly updated based on customer feedback or consultation results. The specialty store server (100) may be configured to link the customized prescription data received from the service provider server (200) with an internal customer relationship management (CRM) system or to provide it directly to the customer through a customer terminal (e.g., app, tablet, etc.). This prescription data consists of a procedure protocol suitable for each customer, product usage instructions, precautions, schedule, etc., and may be provided in the form of a digital report. Furthermore, the specialty store server (100) may record the customer's procedure history and whether the prescription has been fulfilled, and may utilize the data for establishing at least one marketing and customer-customized service strategy, such as managing future visit schedules, tracking effectiveness, and inducing revisits. Through this, the quality of the customer experience can be improved and long-term customer loyalty can be increased. In one embodiment, the specialty store server (100) is linked with at least one of the POS (Point of Sales) system, reservation management system, customer management system (CRM), and treatment record management system of an aesthetic specialty store, and all transaction information and customer activity data generated therefrom are aggregated and stored in real time, and can be transmitted to the service provision server (200) via API or message queue method upon request.

[0041] The service provision server (200) is composed of an analysis data generation unit (210), a prescription data generation unit (220), a storage unit (230), a communication unit (240), and a control unit (250).

[0042] The above analysis data generation unit (210) is composed of a preprocessing module (211), a symptom data generation module (212), and a time series prediction module (213).

[0043] The preprocessing module (211) receives image data containing customer data from the specialty store server (100) and can preprocess the received image data. The preprocessing module (211) can preprocess the image data by performing at least one of normalizing the image resolution, color correction, noise removal, and automatic adjustment of brightness and contrast, and aims to ensure consistency between images acquired from various devices and environments. The preprocessing module (211) can recognize a face or skin area included in the image data and automatically extract a Region of Interest (ROI) that requires analysis. The preprocessing module (211) can apply a face detection algorithm or a YOLO-based object detection technique, and enables precise analysis by segmenting specific areas within the ROI, such as the forehead, cheeks, chin, and around the nose. In addition, the preprocessing module (211) can standardize color information distorted by at least one of external lighting or skin light reflection by performing pixel-based skin tone homogenization processing based on the extracted ROI. This allows for improved accuracy in color-based analysis for detecting skin diseases or abnormal signs. Additionally, the preprocessing module (211) can determine the reliability of the image by analyzing metadata (shooting date and time, device information, etc.) included in the image, or, if necessary, exclude images of quality below a certain standard from the analysis. This enables the performance of a filtering function so that only high-quality data can be used in the subsequent analysis process.

[0044] The symptom data generation module (212) can input the preprocessed image data into an artificial intelligence model capable of multi-symptom classification to extract major symptom data and secondary symptom data. The artificial intelligence model may be composed of at least one of a classification model based on a CNN (Convolutional Neural Network), a deep learning architecture of the ResNet or EfficientNet family, and may be a pre-trained model or a transfer learning model specialized in multi-class classification of skin diseases. The symptom data generation module (212) can extract features from the input image data at the pixel level or patch level, and can identify specific symptoms such as acne, erythema, pigmentation, wrinkles, and loss of elasticity based on the label classification results learned for the corresponding features. Major symptoms are classified as primary symptoms that are most clinically important to the user, and in one embodiment, may be at least one of inflammatory acne, severe pigmentation, and dermal damage. Secondary symptoms are symptoms that are considered relatively less severe or auxiliary, and may include, for example, at least one of enlarged pores, keratin imbalance, or oil / moisture balance abnormalities. Each extracted symptom may be converted into and stored as a data structure containing at least one of the following information: the name (label) of the symptom, location coordinates (x, y, width, height), severity (score or stage), and related probability (probability score). The symptom data generation module (212) can ensure data consistency by considering priority or organizing redundancy between symptoms based on correlation coefficients when multiple symptoms overlap or are interrelated at the same location. Additionally, the symptom data generation module (212) does not stop at simple static extraction, but if history images or analysis data of the same customer from the past exist, it can compare and analyze them to determine whether there has been a change in symptoms and complementarily analyze the current progression of the symptoms based on the trend of change.This allows for the provision of more reliable symptom diagnosis results than analysis results from a single point in time.

[0045] The symptom data generation module (212) can generate comprehensive skin condition data based on the main symptom data and secondary symptom data. The symptom data generation module (212) can generate comprehensive skin condition data that multidimensionally expresses the user's overall skin condition by comprehensively analyzing the extracted main symptom data and secondary symptom data. The comprehensive skin condition data may be composed by quantifying or grading various skin indicators, and may include at least one of the following as main components: skin health score, problem area grade, area requiring improvement, symptom density, symptom distribution map, complex symptom correlation, and potential symptom risk index, but is not necessarily limited thereto. In one embodiment, if 'severe acne' appears intensively on the lower face as a main symptom and is accompanied by excessive sebum and erythema as secondary symptoms, the comprehensive skin condition data is determined to be a "high-risk state for complex acne symptoms due to sebum and inflammatory reaction in the lower face area," and a numerical indicator for this state can be generated and transmitted to the next stage time series prediction module (213). The comprehensive skin condition data described above can be generated through a weighted calculation formula by considering at least one factor among the relationships between symptoms, the intensity of symptoms, the location and extent of occurrence, user history information, and seasonal influence, rather than simply listing individual symptom information. At this time, a correlation matrix, a PCA (principal component analysis)-based dimensionality reduction technique, or an ensemble evaluation model can be utilized. Furthermore, the comprehensive skin condition data is used not only as an input value for predicting future symptom progression in the time series prediction module (213), but also as a basis for quantitative judgment of the current skin condition during the process of generating a treatment protocol. Therefore, the comprehensive skin condition data functions as key summary information representing the overall skin condition, going beyond a summary of symptoms at a single point in time, and can serve as a basis for establishing strategies to improve the user's skin health.

[0046] The time series prediction module (213) can input the comprehensive skin condition data and customer data into a time series prediction model to generate analysis data including information on symptom changes and improvement predictions over a certain period in the future, and transmit it to the specialty store server (100). By predicting changes in the comprehensive skin condition data at specific points along the time axis, the time series prediction module (213) can quantitatively calculate at least one of the timing when a specific skin symptom will be alleviated or worsened, the fluctuation range of the symptom, and the expected period required for improvement. To this end, customer data including at least one of the customer's age, gender, skin type, stress index, menstrual cycle, and seasonal characteristics can be considered together to construct a time series context that is comparable to a similar group with the same symptoms. The time series prediction module (213) can perform time series learning and prediction using at least one of a prediction model based on LSTM (Long Short-Term Memory), GRU (Gated Recurrent Unit), or Temporal Convolutional Network, and can also implement an ensemble method that increases prediction reliability by running multiple prediction models in parallel. The above time series prediction model is trained based on past data of thousands of skin symptoms and treatments, and in one example, if the current intensity of a specific symptom is 70 points, it can output a result in the form that there is an 82% probability that it will be alleviated to 45 points after 2 weeks. In addition, the time series prediction module (213) may include a function to detect early not only the improvement trend over time, but also the possibility of an adverse reaction occurring at a specific point in time and concerns about unexpected symptom aggravation. These prediction results can be used as a basis for adjusting treatment schedules, providing customized notifications to customers, and deciding on protocol changes through the specialty store server (100).In addition, the time series prediction module (213) can generate a multi-layered report in which the prediction results are configured into a visualizable analysis data form, expressed as at least one of an improvement curve for each symptom, a risk indicator heatmap, and an expected time of arrival, which can also be used as customer communication and marketing material.

[0047] In addition, the time series prediction module (213) can produce more reliable results by executing multiple time series prediction models in parallel when generating analysis data and integrating the prediction results of each model in an ensemble manner. The ensemble method may include at least one of a simple average, a weighted average, a majority vote, or a stacking technique, and has the effect of improving overall accuracy by mutually complementing the prediction deviations of prediction models of different structures. Such ensemble-based analysis data can provide the distribution of prediction values ​​by model, thereby helping users and experts intuitively recognize and utilize the uncertainty regarding the skin improvement prediction results. In one embodiment, if the LSTM-based model predicts the possibility of symptom improvement as 78% and the TCN-based model predicts as 83% for the same input data, the system can integrate them to reflect the deviation and reliability of each prediction value in the analysis report along with an ensemble result of 80%. Therefore, the ensemble prediction of the time series prediction module (213) enables the establishment of a more precise treatment plan than a single model-based analysis, and enables a safer and more predictable customized response, especially for high-risk customers or those with complex symptoms.

[0048] The above prescription data generation unit (220) is composed of an optimal treatment timing calculation module (221), a protocol generation module (222), a customized prescription plan generation module (223), and a customized prescription data generation module (224).

[0049] The optimal treatment timing calculation module (221) receives a request from the specialty store server (100) to generate customized prescription data that matches the image data and the analysis data generated by the analysis data generation unit (210), and can extract optimal treatment timing data by analyzing the image data and the analysis data. The optimal treatment timing calculation module (221) can calculate the optimal timing for the start of skin treatment by comprehensively considering the main symptom change prediction information included in the analysis data and the biological and environmental factors included in the customer data. The biological factors considered at this time may include at least one of the customer's menstrual cycle, skin regeneration cycle, and skin sensitivity cycle, and the environmental factors may include at least one of the current season, average humidity, UV index, amount of sunlight, and environmental characteristics of the customer's residential area. In addition, the optimal treatment timing calculation module (221) can identify a stable period in which improvement can begin without symptom aggravation based on the time series prediction results of the analysis data, and can set a candidate group for the start date of treatment by matching the said period with the customer's schedule, lifestyle pattern, and skin reactivity data. Subsequently, the point in time with the highest improvement effect and lowest probability of side effects among the candidate group can be calculated and determined as the optimal treatment time. In one embodiment, for female customers, it may be ideal to start treatment during a stable period of the menstrual cycle when the skin is not sensitive, and certain skin types may show a better treatment response in a mild spring environment than in a hot and humid summer environment. By considering these factors together, it becomes possible to predict the timing of treatment based on multidimensional conditions rather than simple symptom criteria. Additionally, the optimal treatment time calculation module (221) can output numerical data in the form of "the probability of success in a certain percentage if treatment is started in a few days" through a machine learning-based prediction algorithm utilizing past customer data with the same symptoms and similar characteristics.In addition, if the optimal treatment timing calculation module (221) determines that the customer data included in the image data contains information about the customer's future schedule, it may suggest a realistic optimal treatment start date by avoiding conflicts such as the time when the customer's business trip is scheduled or vacation plans are planned, thereby improving both customer satisfaction and treatment continuation rate.

[0050] The protocol generation module (222) can generate at least one optimal treatment protocol by matching the customer data, optimal treatment timing data, and analysis data included in the image data with past cases of the specialty store clinical data. The protocol generation module (222) can design a customized treatment protocol that presents an optimal treatment path by comprehensively analyzing at least one of the major symptom types, symptom intensity, predicted symptom change patterns derived from the analysis data, and skin type, age group, gender, medications being taken, and allergy information included in the customer data. The specialty store clinical data may consist of past performance-based data including at least one of thousands of treatment success / failure cases, symptom-specific improvement rates, side effect incidence rates, and treatment discontinuation rates, and is built in a pre-tagged and normalized form, enabling rapid similarity-based search. Additionally, the protocol generation module (222) can automatically derive multiple treatment protocols with statistically proven effectiveness by comparing and analyzing existing cases with conditions similar to the customer's current state using vectorized feature information. Each derived protocol includes at least one component among a treatment cycle, a step-by-step treatment goal, the type and dosage of the product used, and a procedure cycle, and may be automatically classified as either a non-invasive or invasive approach, particularly reflecting the customer's lifestyle or preferences. Additionally, the protocol generation module (222) enables strategic treatment design beyond simple prescription levels by subdividing step-by-step treatment goals according to the pattern of symptom change during the predicted treatment period and proposing product combinations and management routines that can effectively achieve the goals. In one embodiment, a 'Phase-based treatment protocol' can be configured, such as focusing on soothing and moisturizing for the first 4 weeks and focusing on improving pigmentation and regeneration for the following 8 weeks, thereby inducing efficient treatment progress over time.In addition, the protocol generation module (222) can calculate an overall grade for each protocol by utilizing at least one evaluation indicator among customer satisfaction, revisit rate, and cost-effectiveness, in addition to the existing treatment success rate, and this is transmitted to the subsequent 'customized prescription generation module (223)' to serve as a basis for determining prescription priorities. Therefore, the protocol generation module (222) supports the rapid and scientific generation of individual customer-customized treatment strategies, thereby simultaneously improving the efficiency of medical care at the specialty store and customer satisfaction.

[0051] The customized prescription generation module (223) can calculate a prescription priority by considering at least one of the expected effect, treatment period, cost, and side effect rate among the optimal treatment protocols, and can generate a customized prescription suitable for each protocol. The customized prescription generation module (223) can automatically calculate a priority by applying multidimensional evaluation criteria to each of the multiple optimal treatment protocols derived from the protocol generation module (222). The customized prescription generation module (223) uses at least one of the expected improvement effect (improvement rate), total treatment period, total cost, expected side effect rate, and customer satisfaction-based score as an evaluation indicator, and the evaluation indicators can be quantified through a weighted integrated scoring process. In one embodiment, if there are two protocols that can expect the same effect, the customized prescription generation module (223) can assign a higher priority to the protocol with a lower side effect rate and a shorter treatment period. The customized prescription generation module (223) can clearly distinguish and generate optimal prescriptions according to each protocol assigned a priority, such as in one embodiment, '1st priority: A-7 intensive care for moderate acne', '2nd priority: B-3 mild improvement care', and each prescription can be detailed in at least one of the following: phase configuration (Phase 1, Phase 2, etc.), morning / evening routine, product family and main ingredient, expected effect, usage cycle, and dosage. Additionally, the customized prescription generation module (223) can perform automatic filtering and adjustment of components by considering at least one of the customer's skin type, preferred ingredients, past product usage history, product price range, and whether there is an interaction with medications currently being taken, thereby deriving a realistic prescription that is highly applicable to actual customers.The customized prescription generation module (223) can prioritize product lines available for sale in specialty stores or prioritize the inclusion of specific brand product lines, and can continuously improve the prescription composition algorithm based on subsequent indicators such as customer experience reviews, satisfaction evaluations, and repurchase rates.

[0052] The custom prescription data generation module (224) can generate custom prescription data for each of the above custom prescriptions, including at least one of expected effect, main ingredients, method of use, usage cycle, side effects, and precautions. The custom prescription data generation module (224) can transmit at least one generated custom prescription data to the specialty store server (100). The custom prescription data generation module (224) can convert the components of each custom prescription calculated by the custom prescription generation module (223) into a structured data format and process them into prescription data that can be utilized by the specialty store server (100). In one embodiment, the prescription data may include execution guidelines that include at least one of the treatment period for each phase, a list of products used within the corresponding period, the application site and dosage of each product, the usage cycle (e.g., daily / three times a week), the time of day (morning / evening), and the method of use (including application and absorption time), and may be configured so that the practitioner or customer of the specialty store can easily understand and follow them. In addition, the customized prescription data generation module (224) can automatically summarize and include the efficacy and mechanism of action of the main ingredients included in each prescription, and can also include predicted information regarding expected effects (e.g., 60% improvement rate after 4 weeks), the expected time to improvement, and the duration of the effect. This allows the scientific validity and expected value of the prescription to be clearly conveyed. In addition, the customized prescription data generation module (224) can also generate information that must be notified to the customer in advance. In one embodiment, warnings and precautions such as allergy precautions for specific ingredients, prohibition of concomitant use of retinol, concomitant use of sunscreen, and specific skin irritation reactions can be automatically included, and the incidence rate of expected side effects (e.g., initial dryness 12%) and countermeasures (e.g., recommendation to strengthen moisturization) can also be provided.The customized prescription data generation module (224) can output the generated customized prescription data in at least one format among JSON, XML, or structured HTML, thereby supporting easy utilization in various forms such as an internal system of the specialty store, a mobile app, or a customer guidance report. In one embodiment, the specialty store server (100) can transmit the customized prescription data to a customer mobile app to link it with at least one of a procedure schedule, a product usage reminder notification, or a weekly report. Additionally, the customized prescription data generation module (224) assigns a unique identification number or a history tracking code to the generated prescription data so that it can be utilized for subsequent monitoring and customer feedback analysis, and based on this, it can also contribute to the advancement of prescriptions targeting the same symptom group.

[0053] The storage unit (230) can store all contents of each component of the service provider server (200). The communication unit (240) enables all components of the specialty store server (100) and the service provider server (200) to communicate through a network. The network means may include at least one CDMA-based (or HSDPA-based) mobile communication network, and / or an IEEE 802.16x-based high-speed wireless internet, and / or an IEEE 802.11x-based wireless LAN communication network, but is not necessarily limited thereto. "Network" refers to a connection structure capable of exchanging information between each node, such as terminals and servers, and includes a Local Area Network (LAN), a Wide Area Network (WAN), the World Wide Web (WWW), a wired and wireless data communication network, a telephone network, a wired and wireless television communication network, etc. Examples of wireless data communication networks include, but are not limited to, 3G, 4G, 5G, 3GPP (3rd Generation Partnership Project), LTE (Long Term Evolution), WIMAX (World Interoperability for Microwave Access), Wi-Fi, Bluetooth communication, infrared communication, ultrasonic communication, Visible Light Communication (VLC), and LiFi. The control unit (250) can control all configurations of the service provider server (200).

[0055] The invention has been described with reference to embodiments illustrated in the drawings, but this is merely illustrative, and those skilled in the art will understand that various modifications and equivalent alternative embodiments are possible therefrom. Accordingly, the true technical scope of protection of the invention should be determined by the technical spirit of the appended claims. Explanation of the symbols

[0057] 10; AI-based skin analysis and personalized prescription recommendation system 100; Specialty store server 200; Service provision server 210; Analysis Data Generation Unit 211; Preprocessing Module 212; Symptom data generation module 213; Time series forecasting module 220; Prescription data generation unit 221; Optimal treatment timing calculation module 222; Protocol generation module 223; Customized prescription generation module 224; Personalized prescription data generation module 230; Storage unit 240; Communication unit 250; Control unit

Claims

Claim 1 A method for AI-based skin analysis and customized prescription recommendation comprises: a step in which a specialty server transmits image data containing customer data to a service provider server to request analysis data; a step in which the service provider server generates analysis data based on the image data and transmits it to the specialty server; a step in which the specialty server requests the service provider server to generate customized prescription data corresponding to the image data and analysis data; and a step in which the service provider server generates customized prescription data and transmits it to the specialty server, wherein the step in which the service provider server generates analysis data based on the image data and transmits it to the specialty server comprises: a step in which the service provider server performs preprocessing of the image data; a step in which the service provider server inputs the preprocessed image data into an artificial intelligence model capable of classifying multiple symptoms to extract major symptom data and secondary symptom data; and a step in which the service provider server generates comprehensive skin condition data based on the symptom data. The AI-based skin analysis and customized prescription recommendation method is characterized by further including the step of a service providing server inputting the comprehensive skin condition data and the customer data into a time series prediction model to generate analysis data including information predicting symptom changes and improvements over a future period and transmitting it to a specialty store server, wherein the main symptom data is converted into and stored as a data structure containing at least one information among the name (label), location coordinates (x, y, width, height), severity (score or stage), and related probability (probability score), targeting the main symptom which is the most clinically important primary symptom to the user, and the secondary symptom data is converted into and stored as a data structure containing at least one information among the name (label), location coordinates (x, y, width, height), severity (score or stage), and related probability (probability score), targeting the symptom which has low severity or is considered auxiliary. Claim 2 delete Claim 3 The AI-based skin analysis and customized prescription recommendation method according to claim 1, characterized in that when the service providing server generates analysis data, it executes a plurality of time series prediction models in parallel and integrates the prediction results of each model in an ensemble manner to generate analysis data. Claim 4 A specialty store server that transmits image data containing customer data to a service provider server to request analysis data, and requests the generation of customized prescription data corresponding to the image data and analysis data; and includes a service providing server that generates analysis data based on the image data and transmits it to a specialty store server, and generates customized prescription data and transmits it to the specialty store server. The service providing server performs preprocessing of the image data, inputs the preprocessed image data into an artificial intelligence model capable of classifying multiple symptoms to extract major symptom data and secondary symptom data, generates comprehensive skin condition data based on the symptom data, inputs the comprehensive skin condition data into a time series prediction model to generate analysis data including information predicting symptom changes and improvements over a certain future period, and transmits it to the specialty store server. The major symptom data is converted into and stored as a data structure containing at least one piece of information among the symptom name (label), location coordinates (x, y, width, height), severity (score or stage), and related probability (probability score), targeting the major symptom which is the primary symptom most clinically important to the user. The secondary symptom data is converted into a data structure containing the symptom name (label), location coordinates (x, y, width, height), and related probability (probability score), targeting symptoms with low severity or considered as auxiliary. AI-based skin analysis and personalized prescription recommendation system characterized by being converted into and stored as a data structure containing at least one of information including severity (score or stage) and related probability (probability score). Claim 5 delete Claim 6 In claim 4, the AI-based skin analysis and customized prescription recommendation system is characterized in that, when generating analysis data, the service providing server executes multiple time series prediction models in parallel and integrates the prediction results of each model in an ensemble manner to generate analysis data.

Images