Method and apparatus for providing a comprehensive growth report through the fusion analysis of bone age, lifestyle habits, and growth history data

Abstract

A method and apparatus for providing a comprehensive growth report through the fusion analysis of lifestyle and growth history data as well as bone age data are disclosed. An apparatus for providing a comprehensive growth report according to one embodiment includes: a data collection unit that receives bone age data obtained by analyzing a radiographic bone image through an artificial intelligence model, and receives growth history data and lifestyle data of the subject of the bone image; a preprocessing unit that converts and normalizes the bone age data, growth history data, and lifestyle data according to the same standard and extracts a plurality of time-series features; a growth change prediction unit that inputs the extracted time-series features into an artificial intelligence model to calculate a growth-related predicted risk; and a report generation unit that visually displays the growth-related predicted risk, provides the degree to which the lifestyle data contributed to the calculation of the predicted risk, the change in the predicted risk and the change in predicted growth height according to the change in lifestyle data, and generates a customized recommendation. According to one aspect of the present invention, by fusing static growth data based on bone age with dynamic data obtained through lifestyle and wearable devices, the accuracy and reliability of individual growth prediction can be significantly improved compared to existing simple predicted height calculation methods.

Description

Technology Field

[0001] The present invention relates to a technology for providing a comprehensive growth report including growth prediction, and more specifically, to a method and apparatus for providing a comprehensive growth report through the fusion analysis of bone age data as well as lifestyle habits and growth history data. Background Technology

[0002] Growth prediction technology is a technique designed to predict future changes in height or body shape by analyzing an individual's growth patterns, and it is widely utilized in the growth management and medical diagnosis of children and adolescents. Traditionally, it was common practice to predict growth curves or calculate final predicted height based primarily on bone age imaging results, biometric measurements such as height and weight, and static indicators like age and gender. While this method has the advantage of enabling a certain level of prediction even with relatively simple variables, it had limitations in that it was difficult to adequately reflect growth fluctuations caused by individual lifestyle habits or environmental factors.

[0003] In particular, factors such as sleep duration, exercise frequency, dietary habits, and stress levels are known to have a significant impact on the secretion of growth hormone and physical development, but conventional growth prediction systems have not been able to reflect these acquired factors in real time. Furthermore, despite advancements in IoT sensor technology and wearable devices, research or systems that integrate dynamic biometric data (e.g., activity level, sleep patterns, heart rate, etc.) obtained from sensors into growth prediction models have been limited.

[0004] Therefore, there is a need for technology that can perform personalized growth prediction by fusing static bone age data with dynamic lifestyle data, and further simulate changes in growth indicators or fluctuations in growth risks resulting from behavioral changes.

[0005] Accordingly, the present invention aims to overcome the limitations of conventional technology by proposing a technology for providing a comprehensive growth report that integrates and analyzes bone age and growth-related data with lifestyle habits and IoT sensor data to calculate an individual's future growth (predicted height), quantifies the risk of growth delay or short stature, and visually and quantitatively presents the impact of improvements in sleep duration or exercise habits on growth. The problem to be solved

[0006] According to one embodiment, a method and apparatus for providing a comprehensive growth report are provided by fusion analysis of growth history data including bone age and biometric data, lifestyle data such as sleep, exercise, and eating habits, and real-time biometric data through a wearable device, thereby more precisely predicting an individual's future growth level, quantifying and presenting the risk of growth delay or short stature, and providing feedback by simulating the impact of changes in the user's behavior (lifestyle habits) on growth risk. means of solving the problem

[0007] A comprehensive growth report providing device according to one embodiment includes: a data collection unit that receives bone age data obtained by analyzing a radiographic bone image through an artificial intelligence model, and receives growth history data and lifestyle data of the subject of the bone image; a preprocessing unit that converts and normalizes the bone age data, growth history data, and lifestyle data according to the same standard and extracts a plurality of time-series features; a growth change prediction unit that inputs the extracted time-series features into an artificial intelligence model to calculate a growth-related predicted risk; and a report generation unit that visually displays the growth-related predicted risk, provides the degree to which the lifestyle data contributed to the calculation of the predicted risk, changes in the predicted risk and predicted growth height according to changes in the lifestyle data, and generates a customized recommendation.

[0008] The preprocessing unit includes the growth history data, which contains height, weight, and body mass index (BMI) over time, and converts these values ​​into Z-scores based on the World Health Organization (WHO), the Centers for Disease Control and Prevention (CDC), or Korean growth curves; the lifestyle data includes sleep data over time, is calculated using average sleep time and standard deviation, calculates the difference between bone age and chronological age, and extracts time-series features including growth rate, sleep variability, and exercise variability.

[0009] The growth change prediction unit inputs the extracted time-series features into a time-series artificial intelligence model to calculate growth-related predicted risks, including the risk of short stature and the risk of precocious puberty, and the predicted risks are provided along with their probability values ​​and confidence interval data.

[0010] The data collection unit can detect the above lifestyle data through an activity sensor of a wearable device or smartphone, or collect it from user input through an app.

[0011] The report generation unit provides the risk of short stature, the risk of precocious puberty, future predicted values ​​of growth indicators, the contribution of lifestyle data serving as the basis for the prediction, changes in predicted risk due to changes in lifestyle, changes in predicted growth height, and customized recommendations based on the predicted risk.

[0012] A method for providing a comprehensive growth report according to another embodiment comprises: receiving bone age data obtained by analyzing a radiographic bone image through an artificial intelligence model; receiving growth history data and lifestyle data of the subject of the bone image; converting and normalizing the bone age data, growth history data, and lifestyle data according to the same standard and extracting a plurality of time-series features; inputting the extracted time-series features into an artificial intelligence model to calculate a growth-related predicted risk; and generating a report that visually displays the growth-related predicted risk, provides the extent to which the lifestyle data contributed to the calculation of the predicted risk, changes in the predicted risk and predicted growth height according to changes in the lifestyle data, and provides customized recommendations.

[0013] The steps of transformation, normalization, and extraction of time-series features include the growth history data including height, weight, and body mass index (BMI) over time, and these values ​​are converted into Z-scores based on the World Health Organization (WHO), the Centers for Disease Control and Prevention (CDC), or Korean growth curves; the lifestyle data includes sleep data over time, is calculated using average sleep time and standard deviation, the difference between bone age and chronological age is calculated, and time-series features including growth rate, sleep variability, and exercise variability are extracted.

[0014] The step of calculating growth-related predicted risk involves inputting the extracted time-series features into a time-series artificial intelligence model to calculate growth-related predicted risk, including the risk of short stature and the risk of precocious puberty, and the predicted risk is provided along with its probability value and confidence interval data.

[0015] Lifestyle data can be detected through activity sensors on wearable devices or smartphones, or collected from user input through apps.

[0016] The report generation step provides the risk of short stature, the risk of precocious puberty, future predicted values ​​of growth indicators, the contribution of lifestyle data serving as the basis for the prediction, changes in predicted risk due to lifestyle changes, changes in predicted growth height, and customized recommendations based on the predicted risk. Effects of the invention

[0017] According to one aspect of the present invention, by fusing bone age-based static growth data with dynamic data obtained through lifestyle habits and wearable devices, the accuracy and reliability of individual growth prediction can be significantly improved compared to existing simple predictive height calculation methods. More specifically, the following effects can be obtained.

[0018] First, by reflecting acquired factors such as sleep, exercise, and dietary habits, as well as biosignals in real time, it is possible to dynamically track and predict an individual's growth status. Accordingly, even among users with the same bone age, differences in growth potential based on lifestyle habits can be quantitatively presented.

[0019] Second, by providing quantified risk assessments for growth retardation or short stature, it helps medical professionals or guardians objectively determine priorities for growth management. This goes beyond simple prediction results, laying the foundation for managing growth risks from a preventive perspective.

[0020] Third, by providing a function that simulates fluctuations in growth risk resulting from changes in user behavior, the effects of changes such as increased sleep duration, improved exercise frequency, and better nutrition on future growth can be visually predicted. Through this, users can actively adjust their lifestyle habits to maximize their growth potential.

[0021] Fourth, by linking with medical institutions, growth clinics, school health management systems, etc., it can contribute to the analysis of growth patterns and policy applications not only for individual users but also for groups.

[0022] Accordingly, by expanding existing static prediction-centric growth management technology into a data-fusion, feedback-based intelligent growth management system, it is possible to provide a more personalized and prevention-oriented growth prediction and health management environment. Brief explanation of the drawing

[0023] FIG. 1 is a reference diagram for explaining the concept of a comprehensive growth report providing device according to one embodiment of the present invention, FIG. 2 is a configuration diagram of a comprehensive growth report providing device according to an embodiment of the present invention, FIG. 3 is a flowchart of a method for providing a comprehensive growth report according to an embodiment of the present invention, Figures 4a and 4b illustrate examples of training data for predicting growth changes and result data included in a comprehensive report. Specific details for implementing the invention

[0024] The advantages and features of the present invention and the methods for achieving them will become clear by referring to the embodiments described below in detail together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below but can be implemented in various different forms. These embodiments are provided merely to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the invention, and the present invention is defined only by the scope of the claims. Throughout the specification, the same reference numerals refer to the same components.

[0025] In describing the embodiments of the present invention, if it is determined that a detailed description of known functions or configurations may unnecessarily obscure the essence of the invention, such detailed description will be omitted. Furthermore, the terms described below are defined considering the functions in the embodiments of the present invention, and these may vary depending on the intentions or conventions of the user or operator. Therefore, such definitions should be based on the content throughout this specification.

[0026] Combinations of each block of the attached block diagram and each step of the flowchart may be executed by computer program instructions (execution engine), and since these computer program instructions may be loaded into the processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, the instructions executed through the processor of the computer or other programmable data processing device create a means to perform the functions described in each block of the block diagram or each step of the flowchart.

[0027] Since these computer program instructions may be stored in computer-available or computer-readable memory that can be directed toward a computer or other programmable data processing device to implement a function in a specific way, the instructions stored in said computer-available or computer-readable memory may also be used to produce a manufactured item containing instruction means that perform the function described in each block of a block diagram or each step of a flowchart.

[0028] And since computer program instructions can be loaded onto a computer or other programmable data processing device, instructions that perform a series of operation steps on a computer or other programmable data processing device to create a process executed by a computer and that execute the computer or other programmable data processing device can also provide steps for executing the functions described in each block of the block diagram and each step of the flowchart.

[0029] Additionally, each block or each step may represent a module, segment, or part of code containing one or more executable instructions for executing specific logical functions, and it should be noted that in some alternative embodiments, the functions mentioned in the blocks or steps may occur out of order. For example, two blocks or steps described in succession may actually be performed substantially simultaneously, and the blocks or steps may also be performed in the reverse order of the corresponding functions as needed.

[0030] Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings. However, the embodiments of the present invention exemplified below may be modified in various different forms, and the scope of the present invention is not limited to the embodiments described below. The embodiments of the present invention are provided to more completely explain the present invention to those skilled in the art to which this invention pertains.

[0031] FIG. 1 is a reference diagram for explaining the concept of a comprehensive growth report providing device according to one embodiment of the present invention.

[0032] Referring to FIG. 1, the comprehensive growth report providing device (100) receives data through a plurality of input units, for example, through an AI analysis result (110), a wearable device (120), a smartphone app (130), and a questionnaire (140).

[0033] The AI ​​analysis result (110) calculates the bone age from, for example, an X-ray image of a user's hand and generates growth stage information based thereon. At this time, the AI ​​analysis module extracts morphological features of the hand bone through an image recognition algorithm and compares them with a standard growth database to determine the current bone age and growth rate.

[0034] The technology for determining future growth prediction height through bone age analysis to obtain AI analysis results (110) is described in more detail as follows.

[0035] First, in the GP Atlas-based candidate extraction step, a procedure is performed to present the most visually similar bone age candidates. To this end, the visual similarity between the input image and multiple reference images included in the GP Atlas (a collection of standard bone age images) is calculated using a deep learning-based image comparison algorithm. At this time, the similarity evaluation is centered on morphological features such as the phalanges and carpal bones, which are Regions of Interest (ROIs), and the similarity score of the entire image is calculated by comparing feature vectors for each region. As a result, the system presents the top three bone age candidates with the highest similarity in months and outputs the similarity score for each candidate.

[0036] For example, the top three candidates may be derived as 132 months (similarity 0.91), 120 months (similarity 0.89), and 144 months (similarity 0.88), respectively, and the candidate with the highest similarity among them is selected as the AI ​​analysis result and used in the subsequent prediction key calculation process.

[0037] Next, in the TW3 method-based scoring step, objective and standardized score-based bone age is calculated according to the ossification stage. To this end, the system classifies the maturation stages for 13 major bone regions (e.g., phalanges, radius, ulna, etc.) and assigns scores based on the maturity level of each bone. After calculating the total score by summing the individual scores assigned in this way, the final bone age (in months) is calculated by referring to a standard growth curve (reference curve) corresponding to gender and the total score.

[0038] For example, if a boy's total score is evaluated as 580 points, it can be converted to a bone age of 129 months according to the standard curve.

[0039] In this way, by combining GP Atlas-based visual similarity analysis and TW3 score-based quantitative evaluation, the AI ​​analysis result (110) can become a composite bone age assessment data that simultaneously secures visual intuitiveness and objective reproducibility.

[0040] In addition, by calculating the predicted height (e.g., 166 cm, etc.) together based on each analysis result, the reliability and clinical utility of the final growth prediction report generated by the comprehensive growth report providing device (100) can be improved.

[0041] The wearable device (120) collects the user's biosignals and activity data in real time. For example, data such as sleep time, step count, heart rate, and energy consumption may be included, and these data are transmitted to a comprehensive growth report providing device (100) via Bluetooth or network communication.

[0042] The smartphone app (130) allows the user or guardian to directly input information such as lifestyle habits, meal records, and exercise frequency, or manages data automatically collected by linking with an external device. Additionally, through the app, the user can check their growth status and prediction results in real time.

[0043] The questionnaire (140) is configured to receive qualitative data, such as subjective lifestyle habits, health status, and stress levels, from the user or guardian. This survey data is integrated and analyzed together with AI analysis results and wearable data and reflected in the generation of a growth report.

[0044] Accordingly, the comprehensive growth report providing device (100) fuses and analyzes the aforementioned data to generate an individual's future growth prediction (predicted height), growth risk (risk of low growth, etc.), and growth prediction simulation results based on behavioral changes. The generated results are provided as a comprehensive growth report composed of graphs, numbers, text, etc., and the user can intuitively check their growth status and direction of improvement through this.

[0045] FIG. 2 is a configuration diagram of a comprehensive growth report providing device according to one embodiment of the present invention.

[0046] The comprehensive growth report providing device includes a data collection unit (210), a preprocessing unit (220), a growth change prediction unit (230), and a report generation unit (240).

[0047] The data collection unit (210) receives bone age data analyzed through an artificial intelligence model of a radiographically captured bone image, and receives growth history data and lifestyle data of the bone image subject. Lifestyle data can be collected, for example, through an activity sensor of a wearable device or smartphone, or from user input through an app.

[0048] In other words, the data collection unit (210) is a component that collects growth-related information of the subject in a diverse manner and enables its use in analysis and prediction in subsequent stages. The data collection unit (210) collects data from various sources, including bone age image data, growth history data, lifestyle data, and IoT / wearable sensor data, and is configured to perform missing value correction, noise removal, and normalization processing to ensure data quality during this process.

[0049] To this end, the data collection unit (210) may include a bone age image data analysis module and automatically calculates the bone age (B / A) of the subject by applying an artificial intelligence (AI)-based image analysis algorithm to X-ray or ultrasound images. At this time, the predicted height may also be calculated from the AI ​​analysis results and provided as input data to the growth change prediction unit (230).

[0050] In addition, growth history data can be collected periodically by gathering the subject's height and weight information from home measuring devices, school health checkup systems, or hospital electronic medical record (EMR) systems. The collected data is converted and normalized into standard units and stored in a database.

[0051] Meanwhile, lifestyle data is collected to obtain information related to the subject's acquired growth factors, including sleep, exercise, and dietary information. Sleep information is automatically measured by wearable devices such as smartwatches or obtained through input by guardians or users. Nutrition and dietary information is collected through food logging apps or survey responses, while exercise information is collected in real-time through smartbands, activity trackers on smartphones, or other IoT-based sensors.

[0052] Finally, through IoT or wearable sensors, biological and behavioral data such as the subject's heart rate, activity level, and energy consumption are continuously acquired to dynamically reflect physical activity and physiological state.

[0053] In this way, the data collection unit (210) fusedly secures static (bone age, body measurements) and dynamic (lifestyle habits, IoT sensors) data, thereby providing base data that enables the growth change prediction unit (230) to perform more precise and personalized growth prediction and risk assessment.

[0054] The preprocessing unit (220) converts and normalizes bone age data, growth history data, and lifestyle data according to the same standard and extracts multiple time-series features. More specifically, the growth history data includes height, weight, and body mass index (BMI) over time, and these values ​​are converted into Z-scores based on the World Health Organization (WHO), the Centers for Disease Control and Prevention (CDC), or Korean growth curves. The lifestyle data includes sleep data over time and is calculated as average sleep time and standard deviation. Additionally, the difference between bone age and chronological age is calculated, and time-series features including growth rate, sleep variability, and exercise variability are extracted. The preprocessing unit (220) will be described in more detail below.

[0055] The preprocessing unit (220) is configured to convert and normalize heterogeneous data from different sources input from the data collection unit (210) according to the same standard, thereby generating integrated analysis data that can be utilized in the subsequent growth change prediction stage. In addition, based on the normalized data, various time series features reflecting an individual's growth trend and health status are extracted and provided to the growth change prediction unit (230).

[0056] First, the preprocessing unit (220) performs unit conversion and standardization steps. In this step, body measurement data (height, weight, etc.), biometric data (BMI, etc.), and lifestyle data (sleep, exercise, etc.) are converted into a unified statistical standard to make them comparable.

[0057] For example, height is converted into a Standard Deviation Score (SDS or Z-score) by referring to the WHO growth curve, and the subject's growth level is relatively evaluated based on the value. For instance, SDS=0 means the average level for the age group, SDS=+2 means a level at least 2 standard deviations higher than the average (top approximately 97.5%), and SDS=-2 means a level 2 standard deviations lower than the average (bottom approximately 2.5%). If Z < -2 or Z > +2, the subject is classified as being at risk of short stature or overgrowth.

[0058] In addition, sleep data quantitatively evaluates the stability of sleep patterns by calculating the average sleep time and standard deviation on a weekly basis. For example, the weekly average sleep time is calculated from one week's sleep records, and the consistency and regularity of sleep are assessed by determining the standard deviation. A smaller standard deviation indicates a more consistent and regular sleep pattern, which is utilized as an important variable for growth hormone secretion and growth prediction.

[0059] Subsequently, the preprocessing unit (220) performs a time series feature extraction step. In this step, various time series indicators are calculated to reflect growth trends and lifestyle changes over a certain period. For example, based on height data, the change amount over the last 6 months and 1 year is calculated to calculate the annual growth rate (cm / year), and through this, the normal range and delay of the growth rate are determined.

[0060] In the case of exercise data, the trend of increase or decrease in exercise volume is evaluated by applying linear regression to time series analysis of step count data measured by a wearable device. If the slope (a) is positive, it is interpreted as an increasing trend in exercise volume, indicating high growth potential, and if the slope is negative, it is judged as a decreasing trend in exercise volume, suggesting that it may be a factor inhibiting growth.

[0061] In addition, by calculating the difference between bone age (BA) and chronological age (CA) (BA-CA), the possibility of precocious growth can be quantitatively determined when BA is greater than CA, and the possibility of growth delay can be quantitatively determined when BA is significantly smaller than CA.

[0062] Meanwhile, dietary patterns, the variety, frequency, and quality of nutrient intake are evaluated based on lifestyle surveys or app input. For example, if the nutrition score is in the range of 0.9 to 1.0, it is assessed as a very healthy diet; if it is in the range of 0.6 to 0.8, as a general level or a state requiring improvement; if it is in the range of 0.4 to 0.6, as a diet with a high proportion of processed foods or an unbalanced diet; and if it is less than 0.4, as a nutritional deficiency or irregular diet, and the system is configured to warn of the risk of growth inhibition.

[0063] In this way, the preprocessing unit (220) integrates and normalizes data of various forms and sources and extracts time-series features effective for growth prediction, thereby enabling the growth change prediction unit (230) to perform more accurate and personalized growth prediction and risk analysis.

[0064] Next, the growth change prediction unit (230) inputs the extracted time series features into an artificial intelligence model to calculate a growth-related predicted risk. More specifically, the extracted time series features are input into a time series artificial intelligence model to calculate a growth-related predicted risk, including a risk of short stature and a risk of precocious puberty, and the predicted risk is provided along with the probability value and confidence interval data.

[0065] In other words, the growth change prediction unit (230) is configured to receive data that has been normalized and feature-extracted in the preprocessing unit (220) and to calculate the risk of growth abnormality of the subject through an artificial intelligence-based time series analysis model. The growth change prediction unit (230) comprehensively analyzes time series changes such as individual user's growth data, lifestyle data, and biosignals to predict the likelihood of the occurrence of signs of growth abnormality (e.g., short stature, precocious puberty, etc.) within a certain period in the future.

[0066] Specifically, the growth change prediction unit (230) is configured to learn growth trends and lifestyle change patterns by applying one or more deep learning models, such as LSTM (Long Short-Term Memory), GRU (Gated Recurrent Unit), and TFT (Temporal Fusion Transformer), based on input time series data. These time series models provide higher prediction accuracy than simple statistical-based approaches by considering the cumulative dependency and non-linear correlation of data over time. In this process, the results produced by each model may be combined in an ensemble form to calculate the final risk level. That is, by applying a weighted average of the output values ​​of each model or a voting-based combination method, the bias of a single model is minimized, and the stability and reliability of the growth anomaly prediction are increased.

[0067] The output result of the growth change prediction unit (230) is calculated by classifying it into a risk of short stature and a risk of precocious puberty. Each risk is expressed in the form of a probability score, and a confidence interval is presented along with the prediction result to quantitatively indicate the uncertainty of the prediction. For example, it is output as “risk of short stature: 0.78 (95% confidence interval: 0.71~0.84)” so that medical staff or guardians can intuitively understand the reliability of the prediction result.

[0068] Furthermore, the model can be configured to reflect personalized growth patterns by considering the biological characteristics of the subjects (gender, age, genetic factors, etc.) during training. This allows for the risk level to be assessed differently depending on individual growth rates and lifestyle habits, even for the same absolute numerical change, thereby enabling personalized growth management.

[0069] Accordingly, the growth change prediction unit (230) can be effectively utilized to support clinical judgment and early intervention by calculating the risk of growth abnormality based on a time series AI model, thereby predicting future growth trends beyond a simple current state assessment and early detecting the possibility of growth abnormality.

[0070] Examples of training data and result data will be described later with reference to FIGS. 4a and FIGS. 4b.

[0071] The report generation unit (240) visually displays the growth-related predicted risk level, provides the extent to which lifestyle data contributed to the calculation of the predicted risk level, the change in the predicted risk level and the change in the predicted growth height due to changes in lifestyle data, and generates customized recommendations. More specifically, it provides the risk of short stature, the risk of precocious puberty, future predicted values ​​of growth indicators, the contribution of lifestyle data that serves as the basis for the prediction, the change in the predicted risk level due to changes in lifestyle, the change in the predicted growth height, and customized recommendations based on the predicted risk level.

[0072] In other words, the report generation unit (240) first inputs the preprocessed data into a time-series artificial intelligence model to calculate the probability of growth abnormality risk, and provides a probability value and a confidence interval for each result. By doing so, the uncertainty of the prediction results is expressed quantitatively, thereby supporting medical staff and guardians in making more reliable decisions.

[0073] The generated report is configured to include the following key items.

[0074] 1. Short stature risk items

[0075] The report generation unit (240) calculates the probability of short stature risk based on the case where the height standard deviation score (SDS) of the subject is less than -2, and the risk is displayed along with the median and confidence interval. For example, it is output as “Short stature risk: 72% (95% confidence interval: 64~79%)”.

[0076] 2. Risk of Precocious Puberty Items

[0077] In this section, the difference between Bone Age (BA) and Chronological Age (CA) is calculated to determine the probability of risk of precocious puberty when BA > CA. The results are displayed in the form of probability values ​​and confidence intervals, allowing for a quantitative assessment of whether early growth acceleration is occurring.

[0078] 3. Predicted Value and Uncertainty Information Items

[0079] The report generation unit (240) presents future predicted values ​​of growth indicators (e.g., height SDS, BMI SDS, growth rate, etc.) as median values ​​and quantile ranges (e.g., 10–90%). This visually expresses the uncertainty of the growth prediction, allowing the confidence range of the growth trend to be intuitively understood.

[0080] 4. Lifestyle Contribution Items

[0081] The report generation unit (240) analyzes the proportion of lifestyle factors such as sleep, exercise, and nutrition that contribute to the overall risk and includes the influence of each factor in the report in the form of a percentage or weight. For example, it is displayed as “Sleep: 45%, Exercise: 30%, Nutrition: 25%”.

[0082] 5. Intervention Scenario Simulation Items

[0083] This section presents simulations of risk and predicted adult height changes assuming changes in lifestyle habits. For example, it quantitatively predicts the effects of assumed behavioral changes, such as “increasing sleep time by one hour reduces the risk of short stature from 72% to 55%.” This supports caregivers and clinicians in establishing specific behavioral adjustment plans.

[0084] 6. Recommendation Items

[0085] The report generation unit (240) automatically generates customized recommendations based on the calculated risk level. For example, it is configured to automatically insert recommendation phrases such as “recommendation for regular observation” for low risk, “need for lifestyle improvement” for medium risk, and “need for additional examination or professional medical treatment” for high risk.

[0086] An example of a result report provided including such items is as follows.

[0087] - “Probability of developing short stature within the next 12 months = 30%”

[0088] - “Risk of short stature: 72% (within 12 months)”

[0089] - “Major causes: Slower growth rate over the past year, average sleep of 7 hours or less, BA-CA difference +1.5 years”

[0090] - “Improvement Recommendation: Increasing sleep time by 1 hour reduces risk from 72% to 55% (estimated)”

[0091] In this way, the report generation unit (240) automatically generates a quantitative, explainable AI growth report that includes not only the probability of growth abnormality risk but also the analysis of the cause and improvement simulation, thereby supporting clinicians and guardians in making growth management decisions based on scientific evidence.

[0092] FIG. 3 is a flowchart of a method for providing a comprehensive growth report according to one embodiment of the present invention.

[0093] First, bone age data obtained by analyzing a radiographically captured bone image through an artificial intelligence model is received (310). Then, growth history data and lifestyle data of the bone image subject are received (320). For example, lifestyle data can be detected through activity sensors of a wearable device or smartphone, or collected from user input through an app. The collection of such data is as described in the data collection unit (210) described above with reference to FIG. 2.

[0094] Next, bone age data, growth history data, and lifestyle data are converted and normalized according to the same standard, and multiple time-series features are extracted (330). More specifically, the growth history data includes height, weight, and body mass index (BMI) over time, and these values ​​are converted into Z-scores based on the World Health Organization (WHO), the Centers for Disease Control and Prevention (CDC), or Korean growth curves. The lifestyle data includes sleep data over time, is calculated as average sleep time and standard deviation, the difference between bone age and chronological age is calculated, and time-series features including growth rate, sleep variability, and exercise variability are extracted. This step (330) is identical to the process performed in the preprocessing unit (220) described above with reference to FIG. 2, and is as described in FIG. 2.

[0095] Next, the extracted time series features are input into an artificial intelligence model to calculate the growth-related predicted risk (340). More specifically, the extracted time series features are input into a time series artificial intelligence model to calculate the growth-related predicted risk, including the risk of short stature and the risk of precocious puberty, and the predicted risk is provided along with the probability value and confidence interval data.

[0096] For example, an AI-based time-series analysis model can calculate the risk of growth abnormalities in an analyzed subject. In other words, by comprehensively analyzing time-series changes in individual users' growth data, lifestyle data, and biosignals, it predicts the likelihood of signs of growth abnormalities (e.g., short stature, precocious puberty, etc.) occurring within a certain period.

[0097] Specifically, it is configured to learn growth trends and lifestyle change patterns by applying one or more deep learning models, such as LSTM (Long Short-Term Memory), GRU (Gated Recurrent Unit), and TFT (Temporal Fusion Transformer), based on input time-series data. By considering the cumulative dependency and non-linear correlation of data over time, these time-series models provide higher prediction accuracy than simple statistics-based approaches. In this process, the results produced by each model can be combined in an ensemble form to calculate the final risk level. That is, by applying a weighted average of the output values ​​from each model or a voting-based combination method, the bias of a single model is minimized, and the stability and reliability of growth anomaly predictions are enhanced.

[0098] The output results are calculated by classifying them into the risk of short stature and the risk of precocious puberty. Each risk is expressed in the form of a probability score, and a confidence interval is presented along with the prediction result to quantitatively indicate the uncertainty of the prediction. For example, it is output as “Risk of short stature: 0.78 (95% confidence interval: 0.71~0.84)”, allowing medical staff or guardians to intuitively understand the reliability of the prediction result.

[0099] Furthermore, the model can be configured to reflect personalized growth patterns by considering the biological characteristics of the subjects (gender, age, genetic factors, etc.) during training. This allows for the risk level to be assessed differently depending on individual growth rates and lifestyle habits, even for the same absolute numerical change, thereby enabling personalized growth management.

[0100] Finally, the growth-related predicted risk is visually displayed, the extent to which the lifestyle data contributed to the calculation of the predicted risk, the change in the predicted risk and the change in the predicted growth height according to the change in the lifestyle data are provided, and a customized recommendation is generated (350). More specifically, the extracted time-series features are input into a time-series artificial intelligence model to calculate the growth-related predicted risk, including the risk of short stature and the risk of precocious puberty, and the predicted risk is provided along with the probability value and confidence interval data.

[0101] In addition, it can provide the risk of short stature, the risk of precocious puberty, future predicted values ​​of growth indicators, the contribution of lifestyle data serving as the basis for the prediction, changes in predicted risk due to changes in lifestyle, changes in predicted growth height, and customized recommendations based on the predicted risk.

[0102] The generated report may include, for example, items such as the risk of short stature, the risk of precocious puberty, predicted values ​​and uncertainty information, lifestyle contribution, intervention scenario simulation, and recommendations, and the description of each item is as described above with reference to Fig. 2.

[0103] Figures 4a and 4b illustrate examples of training data for predicting growth changes and result data included in a comprehensive report.

[0104] The growth change prediction unit (230) predicts the growth velocity by utilizing collected lifestyle information and biological data as learning data.

[0105] As illustrated in FIG. 4a, the training data includes sleep, physical activity, and nutritional intake indicators measured over a certain period. Specifically, each data item consists of the average sleep time (sleep_avg_h), variability of sleep time (sleep_var), daily physical activity index (activity_idx), nutritional balance score (nutrition_score), and actual measured growth rate (growth_velocity).

[0106] For example, at time index (time_idx) 24, the first user (A001) is recorded with an average sleep time of 7.5 hours, sleep variability of 0.8, activity index of 1.2, and nutrition score of 0.7, and the growth rate is displayed as unmeasured at that point in time. At the subsequent time index (time_idx) 25, the data for the first user changed to an average sleep time of 8.0 hours, sleep variability of 0.5, activity index of 1.0, and nutrition score of 0.6, and the measured growth rate is 2.6 (cm / month).

[0107] Such time series data is input into a recurrent neural network model based on LSTM (Long Short-Term Memory) or GRU (Gated Recurrent Unit) to learn the effect of changes in lifestyle habits on growth rate. Accordingly, the growth change prediction unit (230) can quantitatively predict changes in growth rate due to changes in acquired factors such as stabilization of sleep patterns or increased physical activity.

[0108] And, the growth change prediction unit (230) quantitatively calculates the growth abnormality risk for each user based on the calculation results of the time series AI model.

[0109] As illustrated in Fig. 4b, the prediction result data consists of information on the risk of short stature (risk_short), risk of precocious puberty (risk_precocious), predicted growth rate (pred_growth_velocity), BMI SDS prediction and major contributing factors, and the confidence interval for each result (confidence_interval).

[0110] For example, the results for the first user at +1 month are calculated as a risk of short stature of 0.12, a risk of precocious puberty of 0.08, and a predicted growth rate of 4.6 (cm / month), and it can be seen that the major contributing factors in this calculation are sleep 40% and exercise 25%.

[0111] This output result is a quantitative representation of the impact that lifestyle changes (e.g., increased sleep time, increased exercise frequency, etc.) reflected in the model's learning process have on the growth rate and the risk of growth abnormalities. Based on the above prediction results, the growth change prediction unit (230) can visually display the trend of changes in the risk of growth abnormalities on a user terminal or a medical institution monitoring system, or automatically provide growth management advice (e.g., improvement of sleep patterns, adjustment of nutritional intake, etc.).

[0112] Here, each risk level is calculated as a probability value, and it can be seen that by including a confidence interval, the uncertainty of the prediction results is specified, thereby improving the reliability of clinical decision-making.

[0113] The present invention has been described above with reference to its embodiments. Those skilled in the art will understand that the present invention may be implemented in modified forms without departing from the essential characteristics of the invention. Therefore, the disclosed embodiments should be considered in an illustrative rather than a restrictive sense. The scope of the invention is defined by the claims, not by the foregoing description, and all variations within the scope of the claims should be interpreted as being included in the invention.

Claims

Claim 1 A comprehensive growth report providing device comprising: an output of a bone age analysis artificial intelligence model that calculates a single bone age and predicted growth height by combining GP Atlas-based image similarity analysis and TW3 score-based ossification evaluation from radiographic bone images, and a data collection unit that receives growth history data of a subject and lifestyle data obtained from a wearable device or user input; a preprocessing unit that converts the growth history data and lifestyle data into Z-scores, mean, standard deviation, and time slopes based on WHO, CDC, or Korean growth curves, and generates a time series feature vector having medical significance including the difference between bone age and chronological age, growth rate, sleep variability, and rate of change in exercise volume; a growth change prediction unit that inputs the time series feature vector into a time series artificial intelligence model to calculate the risk of short stature and the risk of precocious puberty, including probability values ​​and confidence intervals; and a report generation unit that generates a customized intervention recommendation including the probability values ​​and confidence intervals, along with simulated results of changes in risk and predicted growth height due to lifestyle changes. Claim 2 A comprehensive growth report providing device according to claim 1, wherein the preprocessing unit comprises: growth history data including height, weight, and body mass index (BMI) over time, and these values ​​are converted into Z-scores based on the World Health Organization (WHO), the Centers for Disease Control and Prevention (CDC), or Korean growth curves; lifestyle data including sleep data over time, calculated with average sleep time and standard deviation, and the difference between bone age and chronological age is calculated, and time-series features including growth rate, sleep fluctuations, and exercise amount fluctuations. Claim 3 delete Claim 4 A comprehensive growth report providing device according to claim 1, wherein the data collection unit detects the lifestyle data through an activity sensor of a wearable device or smartphone or collects it from user input through an app. Claim 5 A comprehensive growth report providing device according to claim 1, wherein the report generating unit provides a risk of short stature, a risk of precocious puberty, future predicted values ​​of growth indicators, the contribution of lifestyle data serving as the basis for the prediction, changes in predicted risk due to changes in lifestyle, changes in predicted growth height, and customized recommendations based on the predicted risk. Claim 6 A method for providing a comprehensive growth report comprising: the output of a bone age analysis artificial intelligence model that calculates a single bone age and predicted growth height by combining GP Atlas-based image similarity analysis and TW3 score-based ossification evaluation from radiographic bone images, and receiving growth history data of a subject and lifestyle data obtained through a wearable device or user input; the step of converting the growth history data and lifestyle data into Z-scores, mean, standard deviation, and time slopes based on WHO, CDC, or Korean growth curves, and generating a time series feature vector having medical significance including the difference between bone age and chronological age, growth rate, sleep variability, and rate of change in exercise volume; the step of inputting the time series feature vector into a time series artificial intelligence model to calculate the risk of short stature and the risk of precocious puberty, including probability values ​​and confidence intervals; and the step of generating a report that generates a customized intervention recommendation, including the probability values ​​and confidence intervals, and simulated results of changes in risk and predicted growth height due to lifestyle changes. Claim 7 In claim 6, the step of generating the time series feature vector is characterized by the growth history data including height, weight, and body mass index (BMI) over time, and these values ​​are converted into Z-scores based on the World Health Organization (WHO), the Centers for Disease Control and Prevention (CDC), or Korean growth curves; the lifestyle data including sleep data over time, calculated using average sleep time and standard deviation, and the difference between bone age and chronological age is calculated, and time series features including growth rate, sleep variability, and exercise variability are extracted. Claim 8 delete Claim 9 A method for providing a comprehensive growth report according to claim 6, characterized in that the lifestyle data is detected through an activity sensor of a wearable device or smartphone or collected from user input through an app. Claim 10 A method for providing a comprehensive growth report according to claim 6, wherein the report generation step comprises providing a risk of short stature, a risk of precocious puberty, future predicted values ​​of growth indicators, the contribution of lifestyle data serving as the basis for the prediction, changes in predicted risk due to changes in lifestyle, changes in predicted growth height, and customized recommendations based on the predicted risk.

Images