Method for predicting the risk of attention deficit hyperactivity disorder (ADHD) based on dietary quality indicators using machine learning and a method for improving diet using the same

The method uses machine learning to analyze dietary factors to predict ADHD risk and improve dietary quality, effectively reducing ADHD risk through personalized dietary interventions.

KR1020260163705APending Publication Date: 2026-11-17SEOWON UNIV INST OF IND ACADEMY COLLABORATION
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
KR1020250059756
Authority / Receiving Office
KR · KR
Patent Type
Applications
Current Assignee / Owner
Filing Date
2025-05-08
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Current methods lack an effective way to predict the risk of Attention Deficit Hyperactivity Disorder (ADHD) using dietary quality indicators and provide personalized dietary interventions to mitigate this risk.

Method used

A method using machine learning, specifically the AdaBoost classifier, to analyze dietary information such as breastfeeding initiation, vegetable intake, and dietary quality indicators to predict ADHD risk and guide dietary improvements.

Benefits of technology

Enables the prediction of ADHD risk and reduces its likelihood through improved dietary habits, promoting a balanced diet and aiding in early detection and intervention.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure PAT00015_ABST
    Figure PAT00015_ABST
Patent Text Reader

Abstract

The present invention relates to a method for predicting the risk of attention deficit hyperactivity disorder (ADHD) based on dietary quality indicators using machine learning and a method for improving diet using the same. The method comprises: a step of inputting dietary information, which is an input factor for ADHD diagnosis, along with personal information of a child subject to diagnosis, through a user terminal; and a step of performing a machine learning-based ADHD risk prediction on a judgment server using the data input through the user terminal and transmitting the result to the user terminal. The invention also relates to a method for improving diet in which the judgment server guides the direction of dietary habit improvement of the child subject to diagnosis according to the ADHD risk prediction method.
Need to check novelty before this filing date? Find Prior Art

Description

Technology Field

[0001] The present invention relates to a method for predicting the risk of attention deficit hyperactivity disorder (ADHD) based on dietary quality indicators using machine learning and a method for improving diet using the same. More specifically, it relates to a method for predicting the risk of ADHD by evaluating dietary quality in early and late childhood and guiding improvements to improve dietary quality. Background Technology

[0003] Attention Deficit Hyperactivity Disorder (ADHD) is one of the most common neurodevelopmental disorders in childhood, characterized by attention deficit, impulsivity, and hyperactivity. ADHD affects learning ability, social skills, and emotional development, and if left untreated, it can persist into adulthood.

[0004] It is known that the causes of ADHD involve a complex interplay of genetic, neurobiological, and environmental factors.

[0005] In addition, there has been a recent increase in research showing that dietary factors affect children's brain development and behavioral problems.

[0006] In particular, there are reports that the dietary environment in early childhood plays an important role in the development of the nervous system, and that the quality of diet in late childhood is also associated with behavioral and emotional problems.

[0007] At this time, early childhood refers to infancy and toddlerhood up to approximately 6 years of age, and late childhood refers to school-age children from 7 to 12 years of age.

[0008] Meanwhile, machine learning techniques are being widely used in healthcare data analysis recently, contributing to disease prediction and early diagnosis.

[0009] In particular, machine learning capable of simultaneously considering multivariate factors can be a useful tool for predicting the risk of multifactorial disorders such as ADHD. Prior art literature

[0011] Published Patent 10-2023-0153552 The problem to be solved

[0012] Therefore, the objective of the present invention is to provide a method for predicting the risk of attention deficit hyperactivity disorder (ADHD) based on a dietary quality index using machine learning, which is currently used as a useful tool for disease prediction, by evaluating the dietary quality of early and late childhood, and a method for improving diet using the same.

[0013] Meanwhile, it is evident that the technical challenges are not limited to those described above, and that other technical challenges may be derived from the following description. means of solving the problem

[0015] The feature of the present invention for achieving the above-mentioned purpose is a method for predicting the risk of Attention Deficit Hyperactivity Disorder (ADHD) based on dietary quality indicators using machine learning, comprising: a step of inputting dietary information, which is an input factor for ADHD judgment, along with personal information of a child subject to judgment, through a user terminal; and a step of performing a machine learning-based ADHD risk prediction on a judgment server using the data input through the user terminal and transmitting the result to the user terminal.

[0016] The machine learning model above uses the AdaBoost classifier.

[0017] And dietary information uses the timing of breastfeeding initiation, vegetable intake, and dietary quality indicators.

[0018] Another feature of the present invention is a dietary improvement method using a diet quality index-based diet improvement method for attention deficit hyperactivity disorder (ADHD) using machine learning, characterized in that the judgment server guides the direction of dietary habit improvement of the child subject to judgment according to the risk prediction method.

[0019] The above direction for improving dietary habits is characterized by the improvement of the frequency and amount of undesirable eating habits, including the additional intake of vegetables, restriction of sugary foods and beverages, high-fat snacks or fried foods, processed foods including ramen, ham, and sausage, overeating, frequency of visiting fast food restaurants, and use of smartphones during meals. Effects of the invention

[0021] According to the present invention configured as described above, it is possible to predict the risk of attention deficit hyperactivity disorder (ADHD) using dietary data from daily life, and accordingly, it is possible to reduce the risk of attention deficit hyperactivity disorder (ADHD) by improving the diet and also to have the effect of enabling a balanced diet.

[0022] In addition, it has the effect of enabling trained ADHD assessment experts or the general public to easily predict ADHD risk.

[0023] Furthermore, since the effects of the present invention described above are naturally manifested by the composition of the described content regardless of whether the inventor is aware of them, the aforementioned effects are merely a few effects based on the described content and should not be recognized as describing all effects that the inventor has grasped or that actually exist.

[0024] In addition, the effects of the present invention should be further understood from the overall description in the specification, and even if not explicitly stated, if an effect can be recognized as such by a person of ordinary knowledge in the technical field to which the described content belongs through the present specification, it should be considered as an effect described in the present specification. Brief explanation of the drawing

[0026] FIGS. 1 to 3 are drawings showing the correlation between dietary intake measurement and quality index and ADHD symptom score according to the present invention. Figures 4 to 6 are diagrams showing the predictive relationship between dietary quality scores and ADHD symptoms. FIGS. 7 to 11 are drawings illustrating the performance evaluation of a machine learning model for predicting attention deficit hyperactivity disorder (ADHD). FIG. 12 is a diagram showing a system configuration for implementing a method for predicting the risk of attention deficit hyperactivity disorder (ADHD) based on a dietary quality index using machine learning according to the present invention and a method for improving diet using the same. Specific details for implementing the invention

[0027] Below, with reference to the attached drawings, embodiments of the present invention are described in detail so that those skilled in the art can easily implement the invention.

[0028] However, the present invention may be implemented in various different forms and is not limited to the embodiments described herein. In addition, parts unrelated to the description are omitted in the drawings to clearly explain the present invention.

[0029] Throughout the specification, the terms used are merely for describing specific embodiments and are not intended to limit the invention.

[0030] A singular expression includes a plural expression unless the context clearly indicates otherwise.

[0031] Unless otherwise defined in this specification, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the present invention pertains. Terms such as those defined in commonly used dictionaries should be interpreted as consistent with their meaning in the context of the relevant technology and should not be interpreted in an ideal or overly formal sense unless explicitly defined in this specification.

[0033] To explain the present invention, first, the association between dietary factors in early and late childhood and ADHD symptoms and risk is analyzed, the importance of dietary factors in predicting ADHD risk is evaluated through a machine learning-based prediction model, and ADHD risk is predicted using dietary factors.

[0035] [Experimental Method]

[0036] 111 children aged 6 to 11 years were recruited, and the severity and symptoms of ADHD were assessed using the Korean-ADHD Rating Scale (K-ARS), Comprehensive Attention Test, It was evaluated using the CAT), Child Behavior Checklist 6-18, and CBCL 6-18.

[0037] The dietary environment of early childhood (infancy), dietary intake and dietary quality of late childhood were analyzed using various assessment tools such as food records, Nutrition Quotient (NQ), Diet Quality Index-Korean (DQI-K), and Children's Dietary Inflammatory Index (C-DII).

[0038] During this period, the dietary environment of early childhood was investigated, including breastfeeding, formula intake, the timing of the start of complementary feeding, and the use of nutritional supplements before 12 months of age. In late childhood, dietary intake and dietary quality were analyzed using a dietary record, Nutritional Quotient (NQ), the Korean Dietary Quality Index (DQI-K), and the Child Dietary Inflammation Index (C-DII).

[0039] To analyze the relationship between ADHD symptoms and dietary factors, general linear regression analysis and various machine learning models (Logistic Regression, Decision Tree, Support Vector Machine, Random Forest, Gradient Boosting, AdaBoost, XGBoost (eXtreme Gradient Boosting)) were applied.

[0040] To compare the performance of the prediction models, accuracy, precision, recall, F1 score, and AUC (Area under curve) were used as evaluation metrics, and the results for the training, validation, and test sets were examined for each model.

[0041] The analysis of the contribution of major variables was performed using SHAP (Shapley Additive Explanations) analysis.

[0043] [result]

[0044] 1. General characteristics of participants

[0045] Participants were classified into a normal control group (NC, N=58) and a high-risk group (HR, N=53) based on K-ARS ADHD assessment scale scores.

[0046] Specifically, participants aged 6 to 11 were recruited and classified into a normal control group and a high-risk group based on the Korean version of the Attention Deficit Hyperactivity Disorder (ADHD) Rating Scale (K-ARS), and a comprehensive evaluation of dietary factors was conducted for both early childhood and late childhood.

[0047] Early childhood factors included breastfeeding / formula intake history, weaning period, intake of health functional foods, and whether nutritional education or counseling was received.

[0048] The late childhood assessment included dietary behavior, dietary intake (analysis of nutrient and food group intake using a 3-day dietary record method), and dietary quality assessed using the Nutritional Quotient (NQ), the Korean Dietary Quality Index (DQI-K), and the Child Dietary Inflammation Index (C-DII).

[0049] ADHD symptoms were evaluated using the K-ARS, the Comprehensive Attention Test (CAT), and the Child Behavior Checklist (CBCL 6-18).

[0050] For predictive modeling, feature selection and variable contribution analysis were performed using the Recursive Feature Elimination (RFE) and Shapley Additive Explanation (SHAP) methods.

[0051] An ADHD risk prediction model was developed by applying various machine learning algorithms such as Logistic Regression, Decision Tree, Support Vector Machine, Random Forest, Gradient Boosting, AdaBoost, and XGBoost.

[0053]

[0055] A systematic evaluation of clinical characteristics and dietary factors was conducted on the participants.

[0056] As shown in Table 1, participants were classified into a normal control group (NC group) and a high-risk group (HR group) based on their K-ARS scores, and the average K-ARS scores for each group were 8.64 and 22.06, respectively, showing a statistically significant difference (P = 9.64e-18).

[0057] As can be seen in Table 1, the average ages of the NC group and the HR group were 9.23 and 8.24 years, respectively, and there was a significant difference between the two groups in age (P = 0.001).

[0058] However, there were no significant differences between the two groups in gender distribution, body mass index (BMI), birth weight, and family characteristics (household composition, income level, parental education level, smoking status, etc.) (P > 0.05).

[0060] 2. Assessment of the Dietary Environment in Early Childhood (Infancy)

[0061] Dietary status during early childhood is one of the factors that plays an important role in brain function development and mental health.

[0062] Therefore, in the present invention, the dietary environment of infants and toddlers was investigated as shown in Table 2.

[0063]

[0065] The analysis results showed no statistically significant differences between the two groups regarding breastfeeding status and duration, formula intake, timing of complementary food initiation, and use of nutritional supplements before 12 months of age (all P > 0.05).

[0066] However, the age of breastfeeding initiation was significantly higher in the HR group (0.53 ± 0.74 weeks) than in the NC group (0.20 ± 0.41 weeks, P = 0.006).

[0067] In addition, the start time of formula intake was earlier in the HR group (7.21 ± 9.92 weeks) than in the NC group (10.00 ± 13.42 weeks), and the total duration of formula intake was also longer in the HR group (404.62 ± 179.85 days) compared to the NC group (328.25 ± 174.66 days), but there was no significance (P = 0.059).

[0068] This suggests that the HR group tends to introduce formula milk earlier and have a longer period of consumption.

[0069] In addition, significant differences were found in experience with nutrition education or counseling. This appears to have occurred because the survey was conducted on elementary school students, and the difference in the proportion of non-elementary school students between the two groups (HR group: 20.75%, NC group: 3.45%).

[0070] These results suggest that differences in the dietary environment during infancy, particularly the timing of breastfeeding initiation, may be associated with the risk of ADHD.

[0072] 3. Evaluation of Dietary Behaviors and Dietary Intake in Late Childhood

[0073] Table 3 shows the results of comparing dietary behaviors and quantitative and qualitative dietary intake between the NC group and the HR group during late childhood.

[0074]

[0076] Regarding breakfast frequency, the proportion of people in the HR group who ate breakfast 5 to 7 times a week was 75.47%, which was higher than the 58.62% of the NC group.

[0077] On the other hand, the proportion of people consuming it 3 to 4 times a week was 7.55% for the HR group, which was lower than the 27.58% for the NC group.

[0078] However, this difference was not statistically significant (P = 0.055).

[0079] Lunch and dinner frequency were mostly 5 to 7 times per week for both groups, and there was no significant difference (P > 0.05).

[0080] Similarly, there was no significant difference in the proportion of participants who ate with others, including breakfast, lunch, and dinner (Breakfast accompaniment, Lunch accompaniment, Dinner accompaniment; Yes) (all P > 0.05).

[0081] Although the eating out frequency was not statistically significant, the HR group (5.66%) was higher than the NC group (0%) when eating out 5-6 times per week, once a day.

[0082] In terms of the use of dietary supplements (Yes), the HR group was 37.74%, which was significantly higher than the NC group's 15.52% (P = 0.010).

[0083] There was no significant difference between the two groups regarding food security status related to food accessibility, and most responded that they were having a sufficient and varied diet.

[0085] Figures 1 to 3 are diagrams showing the association between dietary intake measurement and quality index and ADHD symptom scores.

[0086] Figures 1 and 2 show the food group intake (%kcal) and dietary quality index scores as the mean (SD) of each group (normal control group (NC) and high-risk group (HR)). P-values ​​were calculated using ordinary least squares (OLS) regression analysis adjusted for age and sex, and statistical significance (P < 0.05) was indicated by an asterisk.

[0087] Figure 3 is a heatmap of the correlation matrix derived from Spearman correlation analysis. The values ​​in the heatmap represent the Spearman correlation coefficients (ρ), and red and blue represent positive and negative correlations, respectively. Significant correlations (P < 0.05) are indicated by yellow boxes.

[0088] In the diagram, the abbreviations C-DII represent the Child Dietary Inflammation Index, DQI-K the Korean Dietary Quality Index, and NQ the Nutrient Index.

[0089] That is, in the present invention, as a result of the analysis of nutrient and food group levels based on a 3-day meal record, there was no significant difference between the two groups in energy and major nutrient intake, as shown in Figure 1.

[0090] However, vegetable intake was significantly lower in the HR group than in the NC group (P = 0.005).

[0091] This difference persisted even after adjusting for age and sex (β = -0.006, P = 0.006).

[0092] This suggests that children at high risk of ADHD consumed fewer vegetables compared to the NC group, regardless of age and gender differences.

[0093] In addition, no significant differences were found in the intake of other food groups, including grains and grain products, potatoes and starch, sugar, legumes, seeds, mushrooms, fruits, meat, eggs, seafood, seaweed, dairy products, oils, tea and beverages, seasonings, and other foods (P > 0.05 in all models).

[0094] Qualitative dietary quality assessment was performed using NQ, DQI-K, and C-DII, which evaluate various aspects of the diet, and as shown in Figure 2, among the three sub-items, the moderation score was significantly lower in the HR group than in the NC group (53.69 ±10.89 vs. 59.24 ±14.05, P = 0.009).

[0095] The moderation domain evaluates whether a person practices a diet that restricts unhealthy sugars, saturated fats, and sodium, as well as undesirable eating behaviors (overeating, frequent use of fast food restaurants, use of smartphones during meals).

[0097] A lower score in this area indicates unhealthy eating habits. Meanwhile, after adjusting for age and gender, no significant differences were observed between the NC and HR groups in balance scores, practice scores, and total NQ scores (all P > 0.05).

[0098] In addition, no significant differences were observed between the two groups in the uncorrected and fully corrected models in the DQI-K total score (NC: 4.80 ±1.67, HR: 4.93 ±1.60) and C-DII total score (NC: 1.27 ±0.71, HR: 1.22 ±0.97) (all P > 0.05).

[0099] In addition, Spearman correlation analysis was performed to examine the association between K-ARS scores, dietary intake, and dietary quality index.

[0100] As a result, as shown in Figure 3, most variables did not show a significant correlation with the K-ARS score, but only vegetable intake (ρ = -0.23, P = 0.022) and NQ-Moderation score (ρ = -0.23, P = 0.017) showed a statistically significant correlation with the K-ARS score.

[0101] This suggests that increased vegetable intake and improved dietary quality are associated with K-ARS scores.

[0103] 4. Association between Dietary Quality Score and ADHD Risk

[0104] We further analyzed the predictive relationship between NQ-Moderation scores and ADHD symptoms.

[0105] To this end, participants were divided into lower quartiles (T1-lowest), middle (T2), and upper (T3-highest) based on NQ-Moderation scores, and regression analysis was performed with the results of the Comprehensive Attention Test (CAT).

[0106] Figures 4 to 6 are diagrams showing the predictive relationship between dietary quality scores and ADHD symptoms.

[0107] In Figure 4, participants were classified into three quartile groups (T1-lowest, T2-middle, T3-highest) based on their NQ-Moderation scores.

[0108] The odds ratios (95% confidence interval, CI) of each index derived from the simple selective attention-visualization test of the Comprehensive Attention Test (CAT) were compared with the T1 and T2 groups and the T3 (control group), and the P-values ​​were calculated using general least squares regression analysis adjusted for age and sex.

[0109] The values ​​inside the bars in Figure 5 represent the number (%) of participants in the normal, borderline, or clinical groups based on predefined cutoff points for each scale (total behavioral problems, internalized problems, externalized problems).

[0110] In addition, it represents the number (%) of participants in the asymptomatic, attention problem and ADHD-related symptoms, or other psychological symptoms groups according to tertiary groups based on predefined cutoff points for the Problem Behavior Syndrome Scale and the Diagnostic and Statistical Manual of Mental Disorders (DSM)-based scales.

[0111] Responses were based on the Child Behavior Checklist 6-18 (CBCL 6-18). P-values ​​were calculated using the chi-square test.

[0112] Statistical significance (p < 0.05) was indicated by an asterisk.

[0113] The scatter plot in Figure 6 shows the relationship between the raw score of the Nutrition Quotient's Moderation domain and the ADHD percentile score among the DSM-based scales of CBCL 6-18.

[0114] Gray dots represent individual data points, the blue line represents the regression line, and the blue shaded area indicates the 95% confidence interval (CI), which represents the range likely to contain the predicted value. The β coefficient and P-value were calculated using standard least squares regression adjusted for age and gender.

[0115] In the diagram, the abbreviations ADHD stands for Attention Deficit Hyperactivity Disorder, CAT for the Comprehensive Attention Test, CBCL for the Child Behavior Checklist, DSM for the Diagnostic and Statistical Manual of Mental Disorders, and NQ for the Nutritional Index.

[0116] As shown in Figure 4, in the visual simple selective attention test of the CAT, the early multiple stimulus response score, an indicator of impulsive behavior, showed a significantly higher risk in the T1 group compared to the T3 group (OR = 3.12, 95% CI: 1.10 - 8.83, P = 0.032).

[0117] In the Auditory Simple Selective Attention Test, the sensitivity index, an indicator of attention discrimination, was significantly lower in the T1 group than in the T3 group (OR = 0.68, 95% CI: 0.48 - 0.96, P = 0.029).

[0118] A comparative analysis between the lower 3rd quartile groups was also conducted in the CBCL 6-18 analysis.

[0119] Figure 5 shows the CBCL 6-18 results classified into total behavioral problems, internalized problems, externalized problems, problem behavior syndrome scales, and DSM-based scales.

[0120] There was no significant difference in the proportion of participants classified as borderline or clinical problems among the three quartile groups on the overall behavioral problems, internalization problems, externalization problems, or problem behavior syndrome scales (all P > 0.05).

[0121] However, a significant difference was observed in the DSM-based scale, and the proportion of participants in the clinical problem category of ADHD was statistically significantly higher in T1 (27.0%) compared to T2 (8.1%) and T3 (10.8%) (P = 0.032).

[0122] In addition, in the fully adjusted model, as shown in Figure 6, a significant negative correlation was confirmed between the NQ-Moderation score and the DSM-based scale for ADHD (ADHD percentile score) (β = -0.578, 95% CI: -0.941 - -0.215, P = 0.002).

[0123] This indicates that the higher the NQ-Moderation score, the lower the ADHD percentile score.

[0124] These results indicate that dietary behaviors, such as low dietary quality scores and an inability to control the intake of sugary foods, high-fat foods, and processed foods, are associated with the prevalence of ADHD in school-aged children (reduced attention and increased risk of impulsive behavior).

[0126] 5. Analysis of the Predictive Relationship Between Dietary Quality Score and ADHD Risk

[0127] As an association between low dietary quality scores and ADHD risk was observed, a predictive modeling approach was used to evaluate the predictive relationship between dietary quality scores and ADHD risk in greater depth.

[0128] Therefore, a machine learning (ML) model for predicting ADHD risk was developed using machine learning techniques, and the predictive power of the dietary quality index and several factors known to be closely related to the onset of ADHD was compared.

[0129] Considering the strengths and limitations of each ML technique, seven ML models were applied and their prediction performance was comprehensively analyzed.

[0130] FIGS. 7 to 11 are drawings showing the performance evaluation of machine learning models for predicting attention deficit hyperactivity disorder (ADHD), and FIGS. 7 to 10 are drawings showing the performance evaluation of seven machine learning models for predicting attention deficit hyperactivity disorder (ADHD) (accuracy per model, F1 score, area under the receiver operation characteristic curve (AUC), and AUC across the entire dataset (training set, validation set, test set).

[0131] Figure 11 shows a list of feature importance values ​​of the optimal prediction model along with Shapley Additive Explanations (SHAP) importance values, indicating that the higher the average value of the SHAP importance, the greater the average influence of the corresponding feature on the model output size.

[0132] In the diagram, the abbreviations AdaBoost stands for Adaptive Boosting, AUC for Area Under the Curve, CCTT for Child Color Line Test, C-DII for Child Dietary Inflammation Index, DQI-K for Korean Dietary Quality Index, NQ for Nutrition Index, SHAP for Shapley Additive Explanations, and XGBoost for Extreme Gradient Boosting.

[0134] A comparison of various machine learning models is summarized in Figures 7 to 9.

[0135] Among the tested models, the AdaBoost model showed the highest performance with an accuracy of 0.65, precision of 0.65, recall of 0.65, F1 score of 0.65, and AUC of 0.70.

[0136] The Decision Tree model achieved an accuracy of 0.65, and although it showed similar performance to AdaBoost with slightly lower precision (0.65), recall (0.65), F1 score (0.65), and AUC (0.63), its discriminative power was lower.

[0137] Other models including Logistic Regression, Support Vector Machine, Random Forest, Gradient Boosting, and XGBoost showed low prediction performance with AUC values ​​ranging from 0.51 to 0.61.

[0139] To evaluate the generalizability of the model, the AUC scores of the training dataset, validation dataset, and test dataset were analyzed.

[0140] In Figure 10, among all models, AdaBoost maintained the highest AUC (0.70) on the test set, followed by Decision Tree (AUC = 0.63) and Gradient Boosting (AUC = 0.61).

[0141] On the other hand, models such as Logistic Regression, Support Vector Machine, and XGBoost had AUC scores on the test set that approached random classification performance (0.57 or less).

[0142] These results confirm that AdaBoost is the highest-performing model for predicting ADHD.

[0143] Next, feature importance was analyzed using SHAP analysis to identify key factors contributing to ADHD prediction.

[0144] Figure 11 shows the SHAP values ​​of the major factors for ADHD in AdaBoost, the model with the highest predictive performance.

[0145] The top 5 most influential traits among all predictor variables included NQ-Balance score (SHAP = 0.18), age (SHAP = 0.17), NQ-Behavior score (SHAP = 0.13), NQ-Abstinence score (SHAP = 0.11), and breastfeeding duration (SHAP = 0.09).

[0146] In particular, among the top five traits, the subcomponents of the Nutritional Quotient (NQ) showed a significant contribution, which highlights the potential impact of poor dietary quality on the risk of ADHD.

[0147] In addition, CCTT2-total test time (SHAP = 0.09), age at which breastfeeding began (SHAP = 0.08), DQI-total score (SHAP = 0.07), and maternal education level (SHAP = 0.03), which measure cognitive processing speed and attention, were identified as related factors.

[0148] These results suggest that indicators related to dietary factors, such as the age of breastfeeding initiation and the quality of school-age diet, have a significant impact on predicting ADHD.

[0150] [Conclusion and Discussion]

[0151] As seen above, children at high risk of ADHD started breastfeeding later, consumed fewer vegetables, and had lower dietary quality (temperance domain) scores than children in the normal control group.

[0152] NQ-temperature scores had a negative correlation with ADHD symptoms, and low dietary quality was associated with impulsivity and reduced attention.

[0153] Among machine learning models, the AdaBoost model showed the highest predictive performance, and dietary quality indicators (NQ scores) and the age of initiation of early breastfeeding were identified as important variables in predicting ADHD risk.

[0154] These results show that improving dietary quality can be an important intervention in the prevention and management of ADHD.

[0155] In particular, it suggests that self-control in food choices and the formation of healthy eating habits can have a positive effect on alleviating ADHD symptoms.

[0156] In addition, machine learning-based predictive models suggest the potential for use in personalized dietary interventions or early screening of at-risk groups, indicating that they can serve as important foundational data for establishing ADHD prevention strategies through dietary and lifestyle interventions in the future.

[0158] Accordingly, we examine the method for predicting the risk of attention deficit hyperactivity disorder (ADHD) based on dietary quality indicators using machine learning according to the present invention, and the method for improving diet using the same.

[0159] To implement the above method, the present invention comprises a user terminal (10) in which personal information of a child subject to judgment and input factors for ADHD judgment, namely the timing of breastfeeding initiation, vegetable intake amount, and dietary quality indicators are input, as shown in FIG. 12; and a judgment server (20) that performs machine learning-based ADHD risk prediction using the data input through the user terminal and transmits it to the user terminal.

[0160] In the above, the user terminal can be any terminal that includes input / output means and communication means, and a dedicated program transmitted from the judgment server is installed to input personal information (name, gender, age, etc.) of the child subject to judgment and to input dietary factor information including the timing of breastfeeding of the child subject to judgment.

[0161] In addition, information such as ADHD risk prediction information output from the above-mentioned judgment server is output.

[0162] The above-mentioned user terminal may be used by the child subject to judgment or their guardian, or by a trained professional judge.

[0163] The above judgment server (20) manages dietary information for each child subject to judgment and performs ADHD risk prediction through machine learning based on the input child subject to judgment information and dietary information, and among machine learning models, uses an AdaBoost classifier to predict the risk.

[0164] The above judgment server (20) records weight information regarding the timing of breastfeeding initiation, vegetable intake, and dietary quality.

[0165] In other words, if breastfeeding starts early after birth, vegetable intake is sufficient during school age, and the quality of the diet is excellent, the risk of ADHD is considered low.

[0167] Afterwards, the risk prediction results are transmitted to the user's terminal, and guidance on improving eating habits is provided based on the dietary information of the child subject to judgment.

[0168] That is, the risk of ADHD is predicted using the timing of breastfeeding initiation, vegetable intake, and dietary quality of the input child being evaluated. The dietary quality is evaluated based on the frequency and quantity of undesirable eating habits, including the restriction of sugary foods and beverages, high-fat snacks or fried foods, and processed foods including ramen, ham, and sausage, as well as overeating, the frequency of visiting fast food restaurants, and the use of smartphones during meals.

[0169] In addition, if the vegetable intake of the child being evaluated is low, vegetable intake is recommended, and additional vegetable intake is entered to verify whether the recommended amount is being consumed, and warnings are provided periodically or intermittently on the user terminal until the input is made, and in the case of dietary quality improvement, warnings are provided on the user terminal whenever undesirable dietary habit data is entered so that dietary habits can be improved.

[0170] Therefore, trained ADHD assessment experts or the general public can easily predict the risk of ADHD and, consequently, easily improve dietary habits, which can significantly reduce ADHD risk factors and promote the health of the child being assessed by adopting a balanced and desirable diet. Explanation of the symbols

[0172] 10 : User terminal 20 : Judgment Server

Claims

Claim 1 A method for predicting the risk of Attention Deficit Hyperactivity Disorder (ADHD) based on dietary quality indicators using machine learning, characterized by comprising: a step of inputting dietary information, which is an input factor for ADHD diagnosis, along with personal information of a child subject to diagnosis, through a user terminal; and a step of performing machine learning-based ADHD risk prediction on a diagnosis server using the data input through the user terminal and transmitting the result to the user terminal. Claim 2 A method for predicting the risk of attention deficit hyperactivity disorder (ADHD) based on dietary quality indicators using machine learning, characterized in that, in claim 1, the machine learning model uses an AdaBoost classifier. Claim 3 A method for predicting the risk of attention deficit hyperactivity disorder (ADHD) based on dietary quality indicators using machine learning, characterized in that, in claim 1, the dietary information includes the timing of breastfeeding initiation, vegetable intake, and dietary quality indicators. Claim 4 A method for predicting the risk of attention deficit hyperactivity disorder (ADHD) based on dietary quality indicators using machine learning, characterized in that, in claim 3, the dietary information further includes the timing of the start of formula feeding. Claim 5 A dietary improvement method using a method for predicting the risk of attention deficit hyperactivity disorder (ADHD) based on dietary quality indicators using machine learning, characterized in that, according to any one of claims 1 to 4, a judgment server guides the direction of dietary habit improvement for the child subject to judgment. Claim 6 A dietary improvement method using a machine learning-based dietary quality indicator-based method for predicting the risk of Attention Deficit Hyperactivity Disorder (ADHD), characterized in that, in claim 5, the direction of dietary habit improvement is the improvement of the frequency and quantity of undesirable dietary habits, including additional vegetable intake, restriction of sugary foods and beverages, high-fat snacks or fried foods, processed foods including ramen, ham, and sausage, overeating, frequency of visiting fast food restaurants, and use of smartphones during meals.