Comprehensive nutrition intervention method and system for children with attention deficit hyperactivity disorder

Abstract

The invention provides a comprehensive nutrition intervention method and system for children with attention deficit hyperactivity disorder, and relates to the technical field of clinical nutrition intervention. According to the method, multi-source information of an ADHD child patient is collected and processed in a standardized mode, object entity data are established, and entity feature description vectors are generated; constructing an evidence-based nutrition knowledge unit library based on a natural language processing technology; analyzing the related knowledge units according to semantic matching, decision tree logic and a regression model, and generating evidence-based hierarchical answers; and personalized nutrition intervention schemes including diet suggestions, nutrient intake demands and behavior and lifestyle management suggestions are output through a natural language generation technology, so that intelligence, personalization and traceability of nutrition intervention are realized.

Description

Technical Field

[0001] This invention relates to the field of clinical nutritional intervention technology, and in particular to a comprehensive nutritional intervention method and system for children with attention deficit hyperactivity disorder. Background Technology

[0002] Attention-Deficit Hyperactivity Disorder (ADHD), commonly known as hyperactivity, is a chronic neurodevelopmental disorder characterized by inattention, hyperactivity, and impulsivity. If ADHD symptoms appear in childhood and remain untreated, they can lead to a series of persistent problems throughout childhood. Its etiology and pathogenesis involve multiple factors, including genetics, neurotransmitter imbalances (such as insufficient dopamine and norepinephrine function), and abnormalities in brain structure and function. Currently, conventional treatments include medication (such as methylphenidate) and behavioral therapy; however, some children do not respond well to medication, experience side effects, or their families seek alternative or complementary therapies. Against this backdrop, nutritional intervention, as a potential adjunctive treatment, is receiving increasing attention and research due to its high safety and ease of implementation. Multiple studies have shown that scientific nutritional intervention has a certain effect on improving ADHD symptoms.

[0003] Existing technologies mostly focus on supplementing single nutrients, dietary patterns, or nutritional compositions. For example, CN119524029B discloses the use of Lactobacillus sanfranciscensis in the preparation of products for the prevention and / or treatment of attention deficit hyperactivity disorder (ADHD), as well as a microbial preparation for the prevention and / or treatment of ADHD. IL234714B discloses an oral dietary supplement composition for alleviating or treating ADHD and related diseases, and its method of use. KR102796157B1 discloses a food composition that helps improve ADHD. Existing technologies suffer from the following shortcomings: fragmented intervention methods: most programs target only one nutritional factor, lacking multi-dimensional integration and making it difficult to comprehensively correct the heterogeneous metabolic abnormalities in children with ADHD; a single evaluation system: mainly relying on behavioral scales, lacking linkage analysis between objective biomarkers (such as blood nutritional indicators, gut microbiota) and behavioral data; low degree of individualization: no dynamic optimization mechanism based on biochemical testing and clinical characteristics has been established, leading to unstable efficacy and poor adherence. Traditional manual nutrition planning suffers from low efficiency, reliance on experience, and difficulty in integrating multi-source data (such as blood indicators, behavioral scales, and gut microbiota). In recent years, AI-based clinical decision support systems have shown potential in personalized nutrition recommendations, but intelligent intervention technologies that integrate multi-dimensional data and achieve real-time interactive optimization for children with ADHD are still lacking. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, the purpose of this invention is to provide a comprehensive nutritional intervention method and system for children with attention deficit hyperactivity disorder (ADHD). By standardizing and integrating multi-source information of ADHD patients, constructing evidence-based medical knowledge in a structured manner, and introducing semantic matching and reasoning mechanisms, the nutritional intervention plan is transformed from experience-based judgment to data-driven, evidence-traceable intelligent decision-making.

[0005] To achieve the above objectives, the present invention provides the following solution:

[0006] A comprehensive nutritional intervention approach for children with attention deficit hyperactivity disorder, comprising:

[0007] A raw dataset of children with ADHD was collected and standardized as comprehensive information for the children to obtain standardized child information data. Based on this standardized child information data, object entity data was established and encoded as entity feature description vectors. The raw dataset included basic information, medical record information, laboratory test data, and lifestyle and dietary intake data.

[0008] Based on medical materials related to ADHD, natural language processing technology is used to screen, extract and label the medical materials to generate nutritional knowledge units containing fields such as knowledge topics, core conclusions or recommendations, evidence levels, evidence sources and related nutrients or foods, in order to build a nutritional knowledge unit library for children with ADHD.

[0009] Input the ADHD patient identifier, retrieve the corresponding entity feature description vector from the object entity data, and use the entity feature description vector as the retrieval input;

[0010] Based on semantic understanding, the entity feature description vector is retrieved and matched with the knowledge units in the ADHD children's nutrition knowledge unit library to obtain a set of knowledge units related to the entity feature description vector, and the priority of the knowledge unit set is determined according to the evidence level and relevance.

[0011] Based on the preset decision tree logic and regression model, the set of knowledge units is comprehensively analyzed to generate an evidence-based hierarchical response that includes intervention conclusions, explanations, levels of evidence, and sources of evidence.

[0012] Natural language processing technology was used to convert the evidence-based stratified responses into personalized comprehensive nutritional intervention plans for the children with ADHD; the personalized comprehensive nutritional intervention plans included dietary recommendations, nutrient intake requirements, and behavioral and lifestyle management recommendations.

[0013] Preferably, the basic information includes the child's age, sex, height, weight, and main clinical manifestations; the medical record information includes past medical history, family history, and behavioral scale scores; the behavioral scale scores include SNAP-IV scale scores and Conners scale scores; the laboratory test data includes continuous blood glucose monitoring results, blood nutrient indicators, food intolerance test results, and gut microbiota data; and the lifestyle and dietary intake include eating habits, exercise habits, sleep patterns, screen time, and daily dietary records.

[0014] Preferably, the lifestyle and dietary intake are collected through a pre-set standardized questionnaire; the standardized questionnaire includes a dietary frequency questionnaire, a 24-hour dietary review questionnaire, and a lifestyle survey questionnaire on sleep, exercise, screen time, and dining environment.

[0015] Preferably, object entity data is established based on the standardized patient information data, and the object entity data is encoded into entity feature description vectors, including:

[0016] The items corresponding to the basic information, medical record information, laboratory test data, and lifestyle and dietary intake in the standardized patient information data are respectively set as entity feature items to obtain the object entity data;

[0017] The object entity data are arranged in a preset order to form the entity feature description vector, so that the entity feature description vector can reflect the basic information, medical record information, laboratory test data and lifestyle and dietary intake of the ADHD child.

[0018] Preferably, during the process of encoding the object entity data into an entity feature description vector, when the SNAP-IV scale score in the medical record information is greater than 40 and the Conners scale T score is greater than 65, the feature "prominent behavioral problems" is added to the entity feature description vector; when the daily added sugar intake in the lifestyle and dietary intake is greater than 50g, the feature "high-sugar diet" is added to the entity feature description vector; when the percentage of erythrocyte Omega3 in the laboratory test data is less than 5%, the feature "Omega3 deficiency" is added to the entity feature description vector; when the serum ferritin in the laboratory test data is less than 30, the feature "insufficient iron reserves" is added to the entity feature description vector; and when the milk IgG test result in the laboratory test data is positive, the feature "dairy intolerance" is added to the entity feature description vector.

[0019] Preferably, each nutrition knowledge unit in the ADHD children's nutrition knowledge unit library also includes a detailed description field, a precaution or potential risk field, and a last update date field, to record the scope of application, implementation method, potential adverse reactions, and update status of each nutrition knowledge item.

[0020] Preferably, the evidence levels include four levels: Level A, Level B, Level C, and Level D. Level A corresponds to evidence formed by multiple high-quality randomized controlled trials or meta-analyses; Level B corresponds to evidence formed by a single randomized controlled trial or multiple cohort studies or case-control studies; Level C corresponds to evidence formed by clinical guidelines, public health guidelines, or expert consensus; and Level D corresponds to evidence formed by observational studies, mechanistic studies, or traditional dietary recommendations. The evidence source records the study type, sample size, main findings, publishing institution, and year of the corresponding study.

[0021] Preferably, the dietary recommendations in the personalized comprehensive nutritional intervention program include a 7-day or 7-to-14-day cyclical diet plan, specific food substitution recommendations, and principles for implementing exclusion diets; the nutrient intake requirements include the daily intake range of protein, fat, and carbohydrates, as well as the micronutrient supplement components and their specific dosage ranges; the behavioral and lifestyle management recommendations include daily or weekly exercise duration, nighttime sleep time, screen time limits, and adjustments to the dining and learning environments.

[0022] Preferably, it further includes:

[0023] After the personalized comprehensive nutritional intervention program has been implemented for a preset period, the original dataset of the ADHD children is re-collected, the standardized child information data and the object entity data are updated, and the updated object entity data is re-encoded into a new entity feature description vector.

[0024] A comprehensive nutritional intervention system for children with attention deficit hyperactivity disorder, comprising:

[0025] The comprehensive information acquisition and entity encoding module for ADHD patients is used to collect raw datasets of ADHD patients, standardize these raw datasets as comprehensive patient information to obtain standardized patient information data, establish object entity data based on this standardized patient information data, and encode the object entity data into entity feature description vectors. The raw dataset includes basic information, medical record information, laboratory test data, and lifestyle and dietary intake data.

[0026] The Evidence-Based Medicine Material Processing and Knowledge Unit Construction Module is used to screen, extract, and label medical materials related to ADHD using natural language processing technology, and generate nutritional knowledge units containing fields such as knowledge topics, core conclusions or recommendations, evidence levels, evidence sources, and related nutrients or foods, in order to build a nutritional knowledge unit library for children with ADHD.

[0027] The ADHD patient feature vector retrieval and preparation module is used to input the ADHD patient identifier, retrieve the corresponding entity feature description vector from the object entity data, and use the entity feature description vector as the retrieval input.

[0028] The semantic matching and knowledge unit retrieval and ranking module is used to retrieve and match the entity feature description vector with the knowledge units in the ADHD children's nutrition knowledge unit library based on semantic understanding, to obtain a set of knowledge units related to the entity feature description vector, and to determine the priority of the knowledge unit set according to the evidence level and relevance.

[0029] The evidence-based stratified analysis and intervention conclusion generation module is used to perform comprehensive analysis on the knowledge unit set based on the preset decision tree logic and regression model, and generate an evidence-based stratified response including intervention conclusion, explanation, evidence level and evidence source;

[0030] The personalized comprehensive nutritional intervention program generation module is used to convert the evidence-based stratified responses into personalized comprehensive nutritional intervention programs for the children with ADHD using natural language generation technology; the personalized comprehensive nutritional intervention programs include dietary recommendations, nutrient intake requirements, and behavioral and lifestyle management recommendations.

[0031] The present invention discloses the following technical effects:

[0032] This invention collects raw datasets of children with ADHD and performs standardized processing to establish object entity data and encodes it into entity feature description vectors. This achieves a systematic integration of basic information, medical records, laboratory test data, and lifestyle and dietary intake, overcoming the shortcomings of existing technologies such as scattered data sources, inconsistent formats, and difficulty in automated analysis. The entity feature description vectors of this invention can comprehensively represent the core characteristics of children with ADHD, providing a calculable and quantifiable basis for subsequent matching and reasoning processes, thereby significantly improving the accuracy and individualization of nutritional intervention decisions.

[0033] This invention constructs a nutritional knowledge unit library based on ADHD-related medical materials. It employs natural language processing technology to screen, extract, and label these materials, solving the problems of low efficiency and incomplete coverage in existing methods that rely on manual review for evidence-based medicine knowledge acquisition. The nutritional knowledge units formed by this invention have structured fields (including knowledge topics, core conclusions or recommendations, evidence levels, evidence sources, and related nutrients or foods), ensuring clear sources and strong evidence for interventions. This makes the recommendation process more reliable, transparent, and compliant with evidence-based medicine standards.

[0034] This invention overcomes the shortcomings of traditional intervention methods, such as the inability to achieve comprehensive reasoning from multi-source information and the tendency to overlook key risk factors, by using entity feature description vectors as retrieval input and performing semantic understanding and matching with a nutritional knowledge unit database for children with ADHD. This semantic matching process not only improves the relevance of knowledge retrieval but also automatically selects the most suitable knowledge units based on the individual characteristics of the child, significantly reducing human judgment errors and shifting the nutritional intervention process from experience-driven to data-driven.

[0035] This invention employs a pre-defined decision tree logic and regression model to comprehensively analyze the matched set of knowledge units, generating an evidence-based hierarchical response that includes intervention conclusions, explanations, levels of evidence, and sources of evidence. This addresses the problems of existing nutritional interventions lacking quantitative reasoning and difficulty in explaining the reasons for recommendations. The reasoning mechanism of this invention can comprehensively judge multiple indicators, providing hierarchical and traceable intervention evidence, thereby significantly improving the scientific rigor and credibility of clinical use.

[0036] This invention utilizes natural language processing (NLP) technology to generate personalized, comprehensive nutritional intervention plans for children with ADHD. The final output includes dietary recommendations, nutrient intake requirements, and behavioral and lifestyle management suggestions, overcoming the significant shortcomings of previous technologies that lacked systematic approach and were difficult to personalize. This invention transforms professional analysis results into understandable and actionable guidance, reducing implementation difficulties for parents and medical staff, and making nutritional intervention plans more feasible and valuable for long-term intervention. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 A flowchart of the method provided in an embodiment of the present invention;

[0039] Figure 2 This is a schematic diagram of the system structure provided in an embodiment of the present invention. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] The purpose of this invention is to provide a comprehensive nutritional intervention method and system for children with attention deficit hyperactivity disorder (ADHD). By constructing entity feature description vectors and combining them with evidence-based knowledge units for intelligent analysis and natural language generation, the method effectively improves the accuracy, individualization, and feasibility of nutritional intervention for children with ADHD, realizing a comprehensive intervention technology that is quantifiable, interpretable, and dynamically adjustable.

[0042] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0043] Figure 1 The method flowchart provided in the embodiments of the present invention is as follows: Figure 1 As shown, this invention provides a comprehensive nutritional intervention method for children with attention deficit hyperactivity disorder, comprising:

[0044] Step 100: Collect the raw dataset of children with ADHD. Standardize the raw dataset as comprehensive information of the children to obtain standardized child information data. Based on the standardized child information data, establish object entity data and encode the object entity data into entity feature description vectors. The raw dataset includes basic information, medical record information, laboratory test data, and lifestyle and dietary intake data.

[0045] Step 200: Based on medical materials related to ADHD, natural language processing technology is used to screen, extract and label the medical materials to generate nutritional knowledge units containing fields such as knowledge topics, core conclusions or recommendations, evidence levels, evidence sources and related nutrients or foods, in order to build a nutritional knowledge unit library for children with ADHD.

[0046] Step 300: Input the ADHD patient identifier, retrieve the corresponding entity feature description vector from the object entity data, and use the entity feature description vector as the retrieval input;

[0047] Step 400: Based on semantic understanding, retrieve and match the entity feature description vector with the knowledge units in the ADHD children's nutrition knowledge unit base to obtain the knowledge unit set related to the entity feature description vector, and determine the priority of the knowledge unit set according to the evidence level and relevance.

[0048] Step 500: Based on the preset decision tree logic and regression model, conduct a comprehensive analysis of the knowledge unit set to generate an evidence-based stratified response that includes intervention conclusions, explanations, levels of evidence, and sources of evidence;

[0049] Step 600: Use natural language processing technology to convert the evidence-based stratified responses into personalized comprehensive nutritional intervention plans for children with ADHD; the personalized comprehensive nutritional intervention plans include dietary recommendations, nutrient intake requirements, and behavioral and lifestyle management recommendations.

[0050] In this implementation, multi-source information is first collected from children with ADHD, including basic information, medical records, laboratory test data, and lifestyle and dietary intake data. This information is then incorporated into a unified raw dataset. Basic information includes age, gender, height, weight, and clinical manifestations; medical records include basic medical records, past medical history, and behavioral scale scores (including SNAP-IV and Conners scores); laboratory test data includes continuous blood glucose monitoring results, blood nutrient indicators (EPA+DHA, ferritin, zinc, magnesium, vitamin D, etc.), food intolerance test results (IgE / IgG), and gut microbiota data; lifestyle and dietary intake data includes daily behavior, sleep habits, eating habits, exercise habits, and daily dietary records. The raw dataset is standardized to eliminate differences caused by different units and formats, providing a unified basis for calculation and analysis. Based on this, object entity data corresponding to the child is constructed, and entity feature description vectors are generated. These vectors comprehensively reflect the child's health status, nutritional status, and lifestyle characteristics, providing a computable input for subsequent evidence-based knowledge matching.

[0051] To generate a medical knowledge structure usable for nutritional intervention reasoning, this implementation method collects clinical nutrition guidelines, public health guidelines, expert consensus, dietary nutrition guidelines, and clinical research literature that meet the requirements of evidence-based medicine. These medical materials are then screened, extracted, and labeled using natural language processing technology to automatically form structured nutritional knowledge units. Each nutritional knowledge unit includes fields such as knowledge topic, core conclusions or recommendations, level of evidence, detailed description, source of evidence, relevant nutrients or foods, precautions or potential risks, and last update date. The level of evidence is set from Level A to Level D according to the reliability of the source, corresponding to different types of evidence-based evidence such as randomized controlled trials, meta-analyses, cohort studies, expert consensus, guidelines, or observational studies. The source of evidence records the study type, sample size, main findings, journal or institution name, and publication year. The resulting nutritional knowledge unit library for children with ADHD ensures that subsequent intervention recommendations have a clear evidence basis and medical interpretability.

[0052] During the intervention plan generation process, the identity identifier of the child with ADHD is input to retrieve the corresponding entity feature description vector. This vector is then used as input for semantic matching to search and analyze the nutritional knowledge unit database, selecting the set of knowledge units most relevant to the child's characteristics. Subsequently, a pre-defined decision tree logic and regression model is used to comprehensively reason about the knowledge unit set, forming an evidence-based hierarchical response, including the intervention conclusion, explanation, level of evidence, source of evidence, and actionable recommendations. Finally, natural language processing technology is used to convert the above evidence-based hierarchical response into a personalized comprehensive nutritional intervention plan for the child, including actionable dietary recommendations (such as recipes and exclusion diets for a certain period), nutrient intake requirements (including dosage requirements for protein, fat, carbohydrates, and specific micronutrients), and behavioral and lifestyle recommendations (including exercise duration, sleep time, and optimization of the dining environment). This implementation method ensures that the intervention results have clear medical evidence, a high degree of individualization, and good operability.

[0053] In one embodiment of the natural language generation of this invention, firstly, based on the entity feature description vector of the ADHD child and the matching set of nutritional knowledge units, several candidate natural language expression sentences are constructed. Each candidate sentence is generated by filling in the core conclusions or suggestions, evidence levels, nutrient names, dosage information, and key points of behavior and lifestyle from the knowledge unit. Subsequently, for each candidate sentence, its semantic relevance score with the child's characteristics and the target intervention topic is calculated. The score for semantic and tonal coherence with the currently generated text. And the score for the degree of evidence support, calculated based on the evidence level and source of the corresponding nutritional knowledge unit. Simultaneously, a length deviation metric is calculated based on the deviation between the statement length and the preset target length. Substitute the above terms into the formula. Perform comprehensive calculations and classify all candidate statements according to... The values ​​are sorted from high to low, and the sentence with the highest score is selected as the output result for the current position, so as to ensure that the content matches the evidence while taking into account semantic coherence and text length control.

[0054] When continuously generating complete personalized nutritional intervention plans, this invention repeatedly applies the aforementioned scoring formula in each round of sentence generation to score and filter candidate statements corresponding to dietary recommendations, nutrient supplementation recommendations, and behavioral and lifestyle recommendations. This ensures that the final output text maintains a high overall relevance to the characteristics of children with ADHD, is highly consistent with the evidence-based basis of nutritional knowledge units, and maintains a natural and coherent narrative style within paragraphs. For example, in generating "daily moderate-intensity exercise..." When given lifestyle recommendations such as "minimum sleep time," "ensuring 9-11 hours of sleep," "eating in a quiet, undisturbed environment with a fixed seat," and "avoiding electronic devices for one hour before meals," scores were increased based on evidence supporting the recommendations in the corresponding knowledge units for exercise, sleep, dining environment, and screen time. At the same time, control the deviation of sentence length from the metric. This makes the overall score Sentences with higher priority are selected into the final solution, thus forming a natural language output text that is both supported by medical evidence and easy for parents to understand and implement.

[0055] In this embodiment, to conduct comprehensive nutritional intervention for children with ADHD, the basic information, medical records, laboratory test data, lifestyle, and dietary intake of the children are first acquired to form a raw dataset. Clinical behavioral assessment includes SNAP-IV scale scores and Conners Parent Scale scores; nutritional intake assessment is obtained through a dietary frequency questionnaire and a 24-hour dietary review; physiological indicators include laboratory test data (complete blood count, trace element testing, and food intolerance testing results); lifestyle information includes sleep patterns, exercise habits, screen time, and dining environment. This raw dataset is then standardized to eliminate dimensional differences and form a structured analytical basis. Based on this, object entity data corresponding to the children is established, and entity feature description vectors are obtained.

[0056] After the entity data is constructed, entity feature description vectors are generated based on the key features in the standardized patient information data for subsequent semantic matching with nutritional knowledge units. In an exemplary encoding method, when the SNAP-IV scale score is greater than 40 and the Conners scale T score is greater than 65, "prominent behavioral problems" is added as a feature to the entity feature description vector; when the daily added sugar intake exceeds 50g, "high-sugar diet" is added to the entity feature description vector; when the red blood cell Omega3 percentage is less than 5%, "Omega3 deficiency" is added to the entity feature description vector; when the serum ferritin level is less than 30, "insufficient iron reserves" is added to the entity feature description vector. For allergy and tolerance-related information, when the milk IgG test result is positive, "dairy product intolerance" is added to the entity feature description vector; when the gut microbiota Shannon index is less than 2.5, "dysbiosis" is added to the entity feature description vector. The entity feature description vectors formed in the above manner can comprehensively reflect the patient's behavioral performance, nutritional status, and lifestyle characteristics.

[0057] The entity feature description vector generated using the above method serves as input for subsequent nutritional intervention inference. This vector, combined with the child's behavioral assessment, dietary intake, laboratory indicators, and lifestyle information, provides a computational feature foundation for retrieving and matching nutritional knowledge units. Because the entity feature description vector includes key features such as behavioral problems, nutritional deficiencies, metabolic risks, and food intolerances, subsequent evidence-based stratified analysis can more accurately identify the child's main risk factors and key areas for nutritional intervention, thus laying the data foundation for the final generation of a personalized comprehensive nutritional intervention plan. This implementation achieves a unified expression of complex information from multiple sources, making the computational logic of nutritional intervention interpretable and scalable.

[0058] In this embodiment, to establish an evidence-based medical knowledge system applicable to nutritional intervention for children with ADHD, an ADHD nutritional medicine database is first established. This involves systematically collecting publicly available clinical nutrition guidelines, public health guidelines, expert consensus, and relevant clinical research literature from both domestic and international sources. The medical materials include, but are not limited to, the ESPGHAN Consensus on Nutritional Management of Childhood ADHD and the American Psychiatric Association's Clinical Practice Guidelines for ADHD. The collected medical materials are then screened, classified, standardized, and coded to eliminate format differences and unify medical expression, resulting in structured medical materials stored in the database. This provides a data source for subsequent automated analysis and knowledge unit generation.

[0059] Based on the aforementioned medical materials, a pre-trained natural language processing model is used to analyze each piece of medical material. This includes processing steps such as topic identification, core information extraction, evidence level assessment, nutrient relevance labeling, and risk warning extraction, enabling each piece of material to form a knowledge record with a unified field structure. This implementation method generates nutritional knowledge units for children with ADHD through automated extraction, allowing the knowledge content to be structured according to information such as topic, core conclusion, evidence level, mechanism of action, evidence source, related nutrients, recommended dosage, contraindications, and update date, thereby improving the accuracy and traceability of knowledge retrieval.

[0060] In one embodiment of the present invention, after collecting, screening, classifying, standardizing, and coding ADHD-related clinical guidelines, expert consensus, and medical literature, each piece of medical material is... The sentence or semantic segment is divided into several parts, and each part is vectorized using a pre-trained language model to obtain the corresponding... Simultaneously, candidate tags for preset knowledge topics or knowledge units are... (For example, the effect of Omega-3 supplementation on ADHD symptoms, the relationship between zinc deficiency and ADHD, the effect of elimination diets on behavior, etc.) Generate vector representations Based on this, through the formula

[0061]

[0062] Calculate the semantic relevance score between the medical material and each knowledge topic, where the first... Importance weight of each segment The prior confidence score is automatically generated based on features such as whether it includes conclusive statements, dosage information, and evidence descriptions. Automatically estimated based on document type and publishing institution; for a given medical material, a relevance score is selected. Topics that are at or above a preset threshold are designated as their respective or strongly related knowledge topics, and materials are organized into corresponding ADHD children's nutrition knowledge units in this way, thereby achieving automatic classification and tag assignment of knowledge units.

[0063] In another embodiment of the present invention, when it is necessary to generate a personalized nutritional intervention plan based on the entity feature description vector of an ADHD child, the entity feature description vector is first mapped into several semantic query fragments, and these query fragments are regarded as "truncate medical materials". Then, it is compared with the topic vectors of each knowledge unit in the ADHD children's nutrition knowledge unit library. Matching is performed. By substituting the query fragment vector into the relevance calculation formula above, the topic of each knowledge unit is obtained. Correlation score with current patient characteristics And combined with prior confidence Automatically adjust the influence weights of different evidence sources. Then follow... The knowledge units are ranked by size, and those with higher scores and higher levels of evidence are prioritized for subsequent decision tree logic judgments and regression model calculations to generate evidence-based tiered answers and personalized comprehensive nutritional intervention plans. This natural language processing formula based on vector representation and weighted cosine similarity enables precise semantic matching between ADHD child characteristics and evidence-based nutritional knowledge without relying on extensive manual rule-making and parameter tuning, thus improving the targeting and reliability of knowledge retrieval and intervention recommendations.

[0064] In one exemplary embodiment, for the topic of "the efficacy of Omega-3 supplementation for ADHD," a structured nutritional knowledge unit can be generated. This knowledge unit includes: the knowledge topic being the impact of Omega-3 supplementation on ADHD symptoms; the core conclusion being that Omega-3 supplements with higher EPA content can improve attention deficit symptoms; an evidence level of A; evidence sources determined based on meta-analysis literature (such as Bloch & Qawasmi, 2011, JAACAP); relevant nutrients including EPA and DHA; relevant food sources including salmon, flaxseed, and algal oil; a recommended dose of 500–1000 mg of EPA daily; contraindications including caution in individuals at risk of bleeding; and an update date of 2024-10-20. This structured knowledge unit can be stored in a unified format in an ADHD children's nutritional knowledge unit database and can be used for subsequent semantic understanding-based retrieval and matching.

[0065] In this embodiment, analysis of medical materials related to Omega-3 fatty acids resulted in a nutritional knowledge unit with structured fields. This knowledge unit documents clinical evidence regarding the improvement of symptoms of inattention, hyperactivity, and impulsivity in children with ADHD, particularly EPA and DHA. Based on findings from multiple high-quality randomized controlled trials and meta-analyses, Omega-3 supplements with higher EPA content have a mild to moderate effect on related symptoms and can be used as an adjunct to comprehensive ADHD management. The knowledge unit further documents the mechanisms of action, including the anti-inflammatory and neuroplasticity-promoting physiological effects of EPA and DHA as important components of nerve cell membranes, and their ability to improve scores on clinical scales such as the ADHD-RS. Evidence sources include the meta-analysis by Bloch & Qawasmi (2011) and the randomized controlled trial by Chang et al. (2018). Relevant nutrients include EPA and DHA, and food sources include deep-sea fish (such as salmon, mackerel, and sardines) and plant sources (such as flaxseed and chia seeds). The precautions note that the effect of this intervention is not consistent for all children, and high doses may cause gastrointestinal discomfort or bleeding risks. It is recommended to use fish oil supplements with quality certification.

[0066] In this embodiment, another knowledge unit is formed based on medical research on the relationship between zinc levels and ADHD symptoms. The topic of this knowledge unit is the association between zinc deficiency and ADHD behavioral manifestations and the potential benefits of supplementation intervention. According to the results of multiple observational studies and some randomized controlled trials, some children with ADHD have low serum zinc levels, and zinc supplementation has a certain adjunctive effect in reducing hyperactivity and impulsivity symptoms, especially in children with zinc deficiency. The knowledge unit documents that zinc is an essential cofactor for many enzymes, participates in the synthesis of neurotransmitters such as dopamine, and plays an important regulatory role in attention. Evidence sources include a review of research on Omega-3 and zinc supplementation by Salehi et al., and a randomized controlled trial by Akhondzadeh et al. on adjunctive treatment with zinc sulfate. This knowledge unit also lists common dietary sources of zinc, such as oysters, red meat, legumes, and whole grains, and suggests that long-term excessive zinc supplementation may lead to copper deficiency or gastrointestinal discomfort; therefore, serum zinc testing is recommended before supplementation.

[0067] Another knowledge unit formed by this implementation relates to the effects of elimination diets on ADHD symptoms. Based on some rigorously designed randomized controlled trials, particularly the INCA study (Pelsser et al., 2011), elimination diets can significantly improve hyperactivity and impulsivity symptoms in some children, especially younger children or those with a history of allergies. The knowledge unit documents the principle behind this method: some children may have intolerances to certain food components (such as artificial colors, preservatives, dairy products, wheat, eggs, etc.), which can trigger behavioral problems. The implementation of elimination diets typically involves consuming only small amounts of hypoallergenic foods for a short period, followed by gradual reintroduction to identify potential trigger foods. Evidence sources include randomized controlled trials and clinical guidelines. The knowledge unit also documents common food categories to avoid and related risk warnings, such as the potential for malnutrition, weight loss, and psychological stress if strictly followed, and therefore must be conducted under professional guidance.

[0068] In this embodiment, the process of determining the nutritional intervention plan for children with ADHD is based on the aforementioned entity feature description vector and the ADHD children's nutritional knowledge unit database. First, by inputting the child's name or unique identifier, the entity feature description vector corresponding to that child is retrieved. This entity feature description vector contains features reflecting behavioral performance, dietary structure, and nutritional status, such as "prominent behavioral problems," "high-sugar diet," "Omega-3 deficiency," and "insufficient iron reserves." Subsequently, this entity feature description vector is used as the retrieval basis to perform semantic association analysis with the ADHD children's nutritional knowledge unit database. Knowledge units in the database that correspond to or are highly relevant to each feature item are searched, forming a preliminary set of matched knowledge units for subsequent evidence-based reasoning and intervention plan design.

[0069] When processing the initially matched set of knowledge units, the sets are sorted and filtered according to the level of evidence, source of evidence, and relevance to the characteristics of the child recorded in each knowledge unit. In an exemplary embodiment, knowledge units with an evidence level of A or B and directly related to the characteristics of the child can be preferentially selected, while retaining some knowledge units with lower evidence levels but supplementary reference value, in order to construct an evidence-based tiered response. This evidence-based tiered response includes at least the intervention conclusion for the main question, an explanation of the basis for the intervention, the corresponding level of evidence and literature sources, and actionable intervention recommendations, enabling medical personnel or nutrition intervention implementers to clearly understand the evidence basis and applicable prerequisites for each recommendation.

[0070] When generating personalized comprehensive nutritional intervention plans, based on the aforementioned evidence-based stratified responses and considering the child's actual situation, the intervention recommendations are organized into several parts, including dietary recommendations, nutritional supplementation recommendations, and behavioral and lifestyle recommendations. Dietary recommendations may include meal plan arrangements and dietary exclusion strategies within a certain period, such as constructing a 7-day cyclical meal plan and reducing the intake of high-sugar and high-additive foods; nutritional supplementation recommendations provide specific nutrient supplement components and their dosage ranges based on the recommended dosages and monitoring indicators recorded in the knowledge unit; behavioral and lifestyle recommendations propose adjustment plans around aspects such as exercise duration, sleep time, and dining environment. Through natural language generation technology, the above structured intervention content can be transformed into a plain, easy-to-understand, and easy-to-implement text output.

[0071] In a specific application example, when the entity characteristic description vector of a child with ADHD includes features such as "prominent behavioral problems," "high-sugar diet," "Omega-3 deficiency," and "insufficient iron reserves," this embodiment can prioritize matching knowledge units related to Omega-3 supplementation and iron reserves. For Omega-3 deficiency, recommendations including algal oil DHA+EPA supplementation can be generated, such as recommending daily supplementation of 200mg DHA and 300mg EPA, with the intervention plan indicating its evidence level as A and the corresponding knowledge unit number for tracing back to relevant literature. For insufficient iron reserves, recommendations for using microencapsulated iron supplements can be given, such as supplementing with elemental iron at 2mg / kg / day, with a suggestion to recheck serum ferritin levels after 4 weeks of supplementation, while also noting that the recommendations are derived from the corresponding knowledge unit. Through this method, personalized comprehensive nutritional intervention plans can achieve a one-to-one correspondence with the child's characteristics, while maintaining evidence-based, interpretable, and actionable content structure.

[0072] In this embodiment, the comprehensive nutritional intervention technology for children with ADHD is first used to acquire and encode multi-source information about the children. Specifically, clinical behavior and lifestyle information is collected through standardized questionnaires, including the SNAP-IV scale, the Conners scale, a dietary frequency questionnaire, and lifestyle questionnaires regarding sleep, exercise, screen time, and eating environment. Simultaneously, laboratory test data, including blood, urine, and stool test results, and electronic medical record information, including diagnosis, medication history, and allergy history, are received. This information collectively constitutes the original dataset of the children with ADHD. After standardization, it is transformed into corresponding object entity data according to a preset feature encoding logic, and further, entity feature description vectors for subsequent analysis are generated, enabling the program to recognize and process the children's behavioral characteristics, nutritional status, and lifestyle characteristics in vector form.

[0073] Building upon this foundation, this embodiment organizes domestic and international clinical guidelines and literature related to ADHD to form an evidence-based basis for nutritional knowledge. Specifically, it collects clinical nutrition guidelines and related research literature, including the ESPGHAN consensus and APA guidelines. These medical materials are collected, screened, classified, standardized, and coded to obtain structured text data. Subsequently, a natural language processing model is used to analyze and label each piece of medical material, automatically generating corresponding nutritional knowledge units. Each knowledge unit is stored using a unified structure, including at least the following fields: knowledge topic, core conclusion or recommendation, level of evidence, detailed description, source of evidence, related nutrients or foods, precautions or potential risks, and last update date. This forms a structured nutritional knowledge unit library for children with ADHD, facilitating subsequent retrieval based on feature search and evidence classification.

[0074] When generating a specific intervention plan, this embodiment retrieves the corresponding entity feature description vector by inputting the child's name or other unique identifier, and combines this vector with an ADHD children's nutrition knowledge unit library. First, based on semantic understanding and matching, the child's feature vector is compared with the knowledge topics, core conclusions, and applicable conditions recorded in the knowledge unit library to retrieve knowledge units related to the child's characteristics, and prioritized according to evidence level and relevance. Then, based on preset decision tree logic judgment and regression model weight calculation, the matched knowledge units are comprehensively analyzed to generate an evidence-based tiered response. The evidence-based tiered response includes at least the intervention conclusion, explanation of the basis for the intervention, evidence level, evidence source, and specific recommendations. On this basis, natural language generation technology is used to transform the above structured output into a natural language description form specific to the child.

[0075] The final personalized comprehensive nutritional intervention plan includes three levels: dietary recommendations, nutritional requirements, and behavioral and lifestyle recommendations. Dietary recommendations may include a 7-14 day cyclical menu, designed considering the child's food preferences and allergies, providing specific implementation plans for exclusion diets, food preparation and cooking suggestions, and guidance on dining out. The nutritional requirements section provides the intake ranges for protein, fat, and carbohydrates, as well as key micronutrient supplements and specific dosage recommendations. The behavioral and lifestyle recommendations section clearly states specific requirements such as 40-60 minutes of moderate-intensity exercise daily, a fixed bedtime and 9-11 hours of sleep, eating in a quiet, undisturbed environment with a fixed seat, and avoiding the use of electronic devices one hour before meals. This invention provides a comprehensive nutritional intervention technology for children with ADHD that organically combines clinical nutrition, evidence-based medicine, and artificial intelligence, ensuring that the intervention plan has a clear evidence base and can be differentiated according to individual characteristics, thus providing truly personalized and scientifically feasible nutritional intervention plans for children with ADHD.

[0076] Corresponding to the above methods, such as Figure 2 As shown, this embodiment also provides a comprehensive nutritional intervention system for children with attention deficit hyperactivity disorder, including:

[0077] The comprehensive information acquisition and entity encoding module for ADHD patients is used to collect raw datasets of ADHD patients, standardize these raw datasets as comprehensive patient information to obtain standardized patient information data, establish object entity data based on this standardized patient information data, and encode the object entity data into entity feature description vectors. The raw dataset includes basic information, medical record information, laboratory test data, and lifestyle and dietary intake data.

[0078] The Evidence-Based Medicine Material Processing and Knowledge Unit Construction Module is used to screen, extract, and label medical materials related to ADHD using natural language processing technology, and generate nutritional knowledge units containing fields such as knowledge topics, core conclusions or recommendations, evidence levels, evidence sources, and related nutrients or foods, in order to build a nutritional knowledge unit library for children with ADHD.

[0079] The ADHD patient feature vector retrieval and preparation module is used to input the ADHD patient identifier, retrieve the corresponding entity feature description vector from the object entity data, and use the entity feature description vector as the retrieval input.

[0080] The semantic matching and knowledge unit retrieval and ranking module is used to retrieve and match the entity feature description vector with the knowledge units in the ADHD children's nutrition knowledge unit library based on semantic understanding, to obtain a set of knowledge units related to the entity feature description vector, and to determine the priority of the knowledge unit set according to the evidence level and relevance.

[0081] The evidence-based stratified analysis and intervention conclusion generation module is used to perform comprehensive analysis on the knowledge unit set based on the preset decision tree logic and regression model, and generate an evidence-based stratified response including intervention conclusion, explanation, evidence level and evidence source;

[0082] The personalized comprehensive nutritional intervention program generation module is used to convert the evidence-based stratified responses into personalized comprehensive nutritional intervention programs for the children with ADHD using natural language generation technology; the personalized comprehensive nutritional intervention programs include dietary recommendations, nutrient intake requirements, and behavioral and lifestyle management recommendations.

[0083] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0084] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A comprehensive nutritional intervention method for children with attention deficit hyperactivity disorder, characterized in that, include: A raw dataset of children with ADHD was collected and standardized as comprehensive information for the children to obtain standardized child information data. Based on this standardized child information data, object entity data was established and encoded as entity feature description vectors. The raw dataset included basic information, medical record information, laboratory test data, and lifestyle and dietary intake data. Based on medical materials related to ADHD, natural language processing technology is used to screen, extract and label the medical materials to generate nutritional knowledge units containing fields such as knowledge topics, core conclusions or recommendations, evidence levels, evidence sources and related nutrients or foods, in order to build a nutritional knowledge unit library for children with ADHD. Input the ADHD patient identifier, retrieve the corresponding entity feature description vector from the object entity data, and use the entity feature description vector as the retrieval input; Based on semantic understanding, the entity feature description vector is retrieved and matched with the knowledge units in the ADHD children's nutrition knowledge unit library to obtain a set of knowledge units related to the entity feature description vector, and the priority of the knowledge unit set is determined according to the evidence level and relevance. Based on the preset decision tree logic and regression model, the set of knowledge units is comprehensively analyzed to generate an evidence-based hierarchical response that includes intervention conclusions, explanations, levels of evidence, and sources of evidence. The evidence-based stratified responses were converted into personalized comprehensive nutritional intervention plans for the children with ADHD using natural language generation technology. The personalized comprehensive nutrition intervention program includes dietary recommendations, nutrient intake requirements, and behavioral and lifestyle management recommendations.

2. The comprehensive nutritional intervention method for children with attention deficit hyperactivity disorder according to claim 1, characterized in that, The basic information includes the child's age, sex, height, weight, and main clinical manifestations; the medical record information includes past medical history, family history, and behavioral scale scores; the behavioral scale scores include SNAP-IV scale scores and Conners scale scores; the laboratory test data includes continuous blood glucose monitoring results, blood nutrient indicators, food intolerance test results, and gut microbiota data; the lifestyle and dietary intake includes eating habits, exercise habits, sleep patterns, screen time, and daily dietary records.

3. The comprehensive nutritional intervention method for children with attention deficit hyperactivity disorder according to claim 1, characterized in that, The lifestyle and dietary intake were collected through a pre-set standardized questionnaire; the standardized questionnaire included a dietary frequency questionnaire, a 24-hour dietary review questionnaire, and a lifestyle survey questionnaire on sleep, exercise, screen time, and dining environment.

4. The comprehensive nutritional intervention method for children with attention deficit hyperactivity disorder according to claim 1, characterized in that, Based on the standardized patient information data, object entity data is established, and the object entity data is encoded into entity feature description vectors, including: The items corresponding to the basic information, medical record information, laboratory test data, and lifestyle and dietary intake in the standardized patient information data are respectively set as entity feature items to obtain the object entity data; The object entity data are arranged in a preset order to form the entity feature description vector, so that the entity feature description vector can reflect the basic information, medical record information, laboratory test data and lifestyle and dietary intake of the ADHD child.

5. The comprehensive nutritional intervention method for children with attention deficit hyperactivity disorder according to claim 2, characterized in that, In the process of encoding the object entity data into an entity feature description vector, when the SNAP-IV scale score in the medical record information is greater than 40 and the Conners scale T score is greater than 65, the "prominent behavioral problems" feature is added to the entity feature description vector. When the daily added sugar intake in the lifestyle and diet exceeds 50g, the "high-sugar diet" feature is added to the entity feature description vector; when the percentage of erythrocyte Omega3 in the laboratory test data is less than 5%, the "Omega3 deficiency" feature is added to the entity feature description vector; when the serum ferritin in the laboratory test data is less than 30, the "insufficient iron reserves" feature is added to the entity feature description vector; when the milk IgG test result in the laboratory test data is positive, the "dairy intolerance" feature is added to the entity feature description vector.

6. The comprehensive nutritional intervention method for children with attention deficit hyperactivity disorder according to claim 1, characterized in that, Each nutrition knowledge unit in the ADHD Children's Nutrition Knowledge Unit Library also includes a detailed description field, a precaution or potential risk field, and a last update date field, to record the scope of application, implementation method, potential adverse reactions, and update status of each nutrition knowledge item.

7. The comprehensive nutritional intervention method for children with attention deficit hyperactivity disorder according to claim 1, characterized in that, The evidence levels are categorized into four levels: Level A, Level B, Level C, and Level D. Level A corresponds to evidence from multiple high-quality randomized controlled trials or meta-analyses; Level B corresponds to evidence from a single randomized controlled trial or multiple cohort studies or case-control studies; Level C corresponds to evidence from clinical guidelines, public health guidelines, or expert consensus; and Level D corresponds to evidence from observational studies, mechanistic studies, or traditional dietary recommendations. The evidence sources record the study type, sample size, main findings, publishing institution, and year of the corresponding study.

8. The comprehensive nutritional intervention method for children with attention deficit hyperactivity disorder according to claim 1, characterized in that, The personalized comprehensive nutritional intervention program includes dietary recommendations such as a 7-day or 7-to-14-day cyclical diet plan, specific food substitution suggestions, and principles for implementing exclusion diets; nutrient intake requirements include daily intake ranges for protein, fat, and carbohydrates, as well as micronutrient supplements and their specific dosage ranges; and behavioral and lifestyle management recommendations include daily or weekly exercise duration, nighttime sleep duration, screen time limits, and suggestions for adjusting the dining and learning environments.

9. The comprehensive nutritional intervention method for children with attention deficit hyperactivity disorder according to claim 1, characterized in that, Also includes: After the personalized comprehensive nutritional intervention program has been implemented for a preset period, the original dataset of the ADHD children is re-collected, the standardized child information data and the object entity data are updated, and the updated object entity data is re-encoded into a new entity feature description vector.

10. A comprehensive nutritional intervention system for children with attention deficit hyperactivity disorder, characterized in that, include: The comprehensive information acquisition and entity encoding module for ADHD patients is used to collect raw datasets of ADHD patients, standardize these raw datasets as comprehensive patient information to obtain standardized patient information data, establish object entity data based on this standardized patient information data, and encode the object entity data into entity feature description vectors. The raw dataset includes basic information, medical record information, laboratory test data, and lifestyle and dietary intake data. The Evidence-Based Medicine Material Processing and Knowledge Unit Construction Module is used to screen, extract, and label medical materials related to ADHD using natural language processing technology, and generate nutritional knowledge units containing fields such as knowledge topics, core conclusions or recommendations, evidence levels, evidence sources, and related nutrients or foods, in order to build a nutritional knowledge unit library for children with ADHD. The ADHD patient feature vector retrieval and preparation module is used to input the ADHD patient identifier, retrieve the corresponding entity feature description vector from the object entity data, and use the entity feature description vector as the retrieval input. The semantic matching and knowledge unit retrieval and ranking module is used to retrieve and match the entity feature description vector with the knowledge units in the ADHD children's nutrition knowledge unit library based on semantic understanding, to obtain a set of knowledge units related to the entity feature description vector, and to determine the priority of the knowledge unit set according to the evidence level and relevance. The evidence-based stratified analysis and intervention conclusion generation module is used to perform comprehensive analysis on the knowledge unit set based on the preset decision tree logic and regression model, and generate an evidence-based stratified response including intervention conclusion, explanation, evidence level and evidence source; The personalized comprehensive nutritional intervention program generation module is used to convert the evidence-based stratified responses into personalized comprehensive nutritional intervention programs for the children with ADHD using natural language generation technology. The personalized comprehensive nutrition intervention program includes dietary recommendations, nutrient intake requirements, and behavioral and lifestyle management recommendations.

Images