Integrated Framework for Managing Childhood Obesity Based on Biobanks, AI Tools and Methods, and Serious Games

Abstract

The growing epidemic of childhood obesity is a major threat to their overall development and poses a number of challenges for health systems. We propose an integrated framework to comprehensively address childhood obesity. The proposed architecture addresses essential data management and pre-processing functionalities to support scalable, secure, and privacy-preserving data processing in distributed environments. We are also incorporating a health data-driven AI approach for predictive analytics and decision support. There is additionally a User Engagement Layer, which serves as the main point of interaction for users. It connects individuals to system capabilities, facilitating data collection, progress monitoring, and insights. Finally, we present four serious games designed to address protective factors (such as physical activity and healthy eating) and mitigate risk factors (such as excessive screen time and unhealthy food choices). The identified educational objectives were translated into game elements including goal setting, social support, and positive reinforcement. In order to facilitate our approach, we have described the essential data flows and user interactions within our Biobank architecture.

1. Introduction

Childhood obesity is a critical public health challenge with far-reaching consequences for physical, mental, and social well-being [1]. As the prevalence of childhood obesity continues to increase worldwide, particularly in developed nations, it has become a major focus for researchers and healthcare practitioners alike [2,3]. The root of childhood obesity are complex and multifactorial, involving genetics, lifestyle behaviors, environmental influences, and socioeconomic factors [4]. Addressing these issues requires comprehensive and innovative solutions that go beyond traditional approaches, incorporating technological tools to engage children, families, and healthcare providers in meaningful ways.

In recent years, advances in artificial intelligence (AI), data analytics, and digital health tools have introduced new opportunities to tackle childhood obesity more effectively [5]. AI can enhance our understanding of the patterns and risk factors associated with childhood obesity through predictive modeling and personalized intervention. Serious games, which use game mechanics to motivate the change in healthy behaviors, offer an engaging approach to educating children and promoting healthier lifestyles. In addition, Biobanks, which collect and store biological data, provide invaluable resources for analyzing genetic and metabolic factors that influence the risk of obesity [6]. Integrating these elements into a unified framework could revolutionize the prevention and management of childhood obesity, paving the way for personalized, effective interventions [7].

Although there are numerous technologies to combat childhood obesity, they are often implemented in isolation, limiting their potential impact. AI tools, serious games, and Biobanks each contribute uniquely to understanding and managing obesity, but lack cohesive integration that could maximize their benefits. Current obesity interventions also face challenges with sustained engagement, particularly among children, and there are gaps in leveraging genetic data to tailor interventions. Furthermore, public acceptance and data privacy concerns add complexity to implementing these technologies in healthcare. Thus, a comprehensive, integrated framework that unites Biobanks, AI tools, and serious games is needed to enhance user engagement, personalization, and clinical efficacy in combating childhood obesity.

To the best of our knowledge, we premiere an architecture that integrates Biobanks, AI tools, and serious games in a unified framework for childhood obesity management. The contribution of this work is summarized as follows:

• We propose a modular architecture that seamlessly integrates data from Biobanks, processes it through AI-driven analytics to assess obesity risk and recommend interventions, and deploys innovative applications to reinforce healthy behaviors in children. This architecture allows for a comprehensive approach, where health, behavioral, and environmental factors are all addressed within a single cohesive system.
• We implement a suite of serious games, which are designed to increase children’s nutritional knowledge and promote physical activity in an interactive and sustainable manner.
• We present architectural data-flow diagrams illustrating the information exchange between system components. These diagrams provide insight into how data and computational processes move through the framework architecture and how users interact with different elements of the architecture.

The remainder of this paper is structured as follows. Section 2 provides a review of the current literature on childhood obesity interventions, Biobanks, AI tools and methods, and serious games, outlining the existing gaps and the potential for integrated solutions. Section 3 introduces the proposed architecture, detailing each component and how they interact within the system. In addition, we describe the serious games developed as part of the framework, including their design principles, educational objectives, and technical specifications. In Section 4, we present a series of sequential diagrams that illustrate the essential information pathways and user engagement processes embedded within our Biobank framework. Finally, Section 5 concludes the paper with a summary of key findings, discusses limitations, and proposes directions for future research.

2. Evidence-Based Behavioural Modification for the Prevention of Childhood Obesity

2.1. Related Work

2.1.1. Childhood Obesity

Childhood overweight/obesity is a multifactorial disease, caused by a dynamic interplay between several factors related to the “specific external exposome” (diet, physical activity and sleep), those related to the “general external exposome” (perinatal exposures and social/built environment), and those related to the “internal exposome” (genetics/epigenetics and metabolomics) [8,9,10]. All these factors directly or indirectly affect the energy equilibrium, defined as the balance between energy intake (from foods/beverages) and energy expenditure (from basal metabolic rate, exercise-related energy expenditure and diet-induced thermogenesis) [11]. Although the energy equilibrium must be positive to support normal physical growth during childhood, a great imbalance in favour of energy intake can lead to excess body weight and the development of overweight/obesity in the long term.

Among the aforementioned risk factors, lifestyle habits, most importantly diet and physical activity, are largely modifiable and thus are key targets of interventions against childhood overweight/obesity.

Various components of dietary habits have been investigated in the development of childhood overweight/obesity, in particular food and food-group intake [12], as well as adherence to dietary patterns (e.g., Mediterranean diet, Western diet) [13] and meal patterns (e.g., eating breakfast, having dinner, the frequency and type of snacking) [14]. Current evidence points towards a significant detrimental impact of sugar-sweetened beverages and fast foods on overweight/obesity, whereas the role of refined grains and meat products remains less evident [12]. Moreover, a higher adherence to prudent/healthy dietary patterns, which consist of high intakes of fruits, vegetables, whole grains, fish, nuts, legumes, and yogurt, as well as low intakes of sugar and animal fat, has been associated with decreased odds of overweight/obesity, whereas a Western-type dietary pattern, which is characterised by sugary foods and drinks, processed foods, fast food, animal products, and refined grains, has been positively associated with markers of obesity [13]. Overall, diet quality seems to be a significant determinant of childhood overweight/obesity [15]. In relation to meal patterns, regular breakfast consumption has been associated with lower body mass index and reduced odds of overweight/obesity, whereas breakfast skipping may be a detrimental factor [16]. Besides breakfast, there is also some evidence to support that regular family meals are protective towards obesity, highlighting a potential role of the family environment in eating patterns [16].

In light of the above, the most recent guidelines from the American Academy of Pediatrics (AAP) recommend reducing sugar-sweetened beverages, avoiding breakfast skipping, and consuming fruits and vegetables for 5 days as the most effective behavioral strategies for treating childhood overweight/obesity [17].

Besides diet, physical activity is another important determinant of body-weight status, being a main contributor to energy expenditure, and health in general. A high level of physical activity has been associated with improvements in several health outcomes, including bone health, fitness, cardiometabolic profile, cognitive function, and mental health, both in adult and youth populations [18]. Focusing on body-weight status, the available epidemiological research has consistently shown that engagement in physical activity, particularly of moderate-to-vigorous intensity, is inversely associated with various adiposity-related outcomes [19], and conversely that sedentariness, particularly increased time spent in front of screens for recreation, is associated with increased adiposity [20], in children’s cohorts. The isotemporal substitution of sedentary time with physical activity has also been associated with lower adiposity indices among the youth, suggesting that increasing time spent in physical activity at the expense of sedentary activities might represent the optimal approach to prevent childhood overweight/obesity [21]. Besides epidemiological data, recently published systematic reviews and meta-analyses of clinical trials, mostly implemented at the school setting, have concluded that physical activity, either alone or as part of multicomponent lifestyle interventions, is an effective strategy for the prevention of childhood overweight/obesity [22,23,24].

In light of the aforementioned evidence, the 2020 guidelines of the World Health Organization (WHO) recommend a minimum of 60 min of moderate-to-vigorous physical activity per day for children and adolescents in order to achieve these health benefits [25]. Accordingly, the 2020 position statement of the European Childhood Obesity Group and the European Academy of Pediatrics acknowledges physical activity as a significant determinant of body-weight status and provides recommendations for adopting a physically active lifestyle, similar to those of WHO, that can contribute to the prevention of excessive body weight in childhood and adolescence [26]. However, based on a 2020 synthesis of 298 surveys from 146 countries, 81.0% of 11–17-year-olds (77.6% of boys and 84.7% of girls) do not meet these recommendations, highlighting the urgent need for national and global actions to promote physical activity and reduce sedentariness among children and adolescents [27].

2.1.2. AI in Healthcare

Artificial intelligence (AI) is increasingly being integrated into healthcare, improving predictive models and enabling personalised interventions. AI technologies can help address issues such as childhood obesity through real-time data analysis, risk-assessment models and tailored healthcare plans.

A key application of AI in healthcare is predictive modelling, where algorithms analyse historical healthcare data to predict future trends or a patient’s evolution. In the case of childhood obesity, there are multiple factors such as genetics, lifestyle and socio-economic indicators [28] that need to be taken into account by these models.

Various algorithms have been used to predict childhood obesity, such as Decision Trees to predict obesity over the 2–10 year age range [5], Random Forest and Gradient Boosting to predict malnutrition risk [29], and Neural Networks and Support Vector Machines have been investigated for their potential to handle complex, non-linear relationships in obesity data [30]. Machine learning models can accurately predict obesity risk in children as young as two years old [31].

AI can enable personalized intervention strategies tailored to individual genetic profiles, environmental factors, and behavioral patterns [32]. It has the potential to identify early indicators of obesity and enable targeted interventions. The integration of AI with EHRs provides clinicians with real-time information on patient progress and enables the development of tailored treatment plans [33]. Using prognostic data, AI can tailor dietary recommendations [29], exercise regimes and psychological support mechanisms, ensuring that interventions are closely tailored to individual needs and increasing their effectiveness.

To support such an AI-driven approach, access to diverse and high-quality datasets is essential. Real-world health data are often subject to strict privacy protections and regulatory restrictions, which are essential to protect patients. In this context, the use of synthetic data has emerged as a powerful alternative, enabling the training and validation of AI models without compromising patient confidentiality [34]. Research papers have demonstrated that synthetic datasets can maintain statistical similarity to source data while effectively eliminating re-identification risks through methods such as differential privacy and noise injection during data generation [35]. These datasets have been successfully employed in model training, with negligible performance differences compared to models trained on real data [36].

2.1.3. Biobanks

A Biobank is a repository that collects, stores, and manages biological samples, such as blood, tissue, and DNA, and associated health and genetic data [37]. The main purpose of Biobanks is to facilitate scientific research, to help scientists understand disease, to use data to uncover complex genetic relationships to improve diagnosis, and to complement research with practical applications [38,39].

Modern Biobanks depend on IT infrastructures capable of hosting large and heterogeneous datasets. For example, the German Biobank Alliance has implemented a federated IT infrastructure to manage and harmonise Biobank data across institutions [40].

Biobanks can be used to identify the factors that lead to disease, both genetic and acquired. In the Netherlands, a Biobank has been established to provide ongoing information on the progression and treatment of newly diagnosed patients with type 2 diabetes, with a focus on personalised treatment [41].

Despite their vital contribution, Biobanks face many challenges. First, they operate under different regulatory, ethical, and operational frameworks, making data sharing and collaboration complex. Another challenge is the management and analysis of large, heterogeneous datasets. This involves integrating genetic, phenotypic, and clinical data, while ensuring data security and the privacy of participants. Building trust is essential to ensure that this is shared and supported by the population at large.

2.1.4. Serious Games

Game-based health interventions produce small but significant BMI reductions in overweight/obese youth [42], with multicomponent approaches showing greater effectiveness than standalone interventions. Serious games show promise as an educational strategy to improve knowledge and encourage behavioral changes in overweight or obese children [43]. Serious games can tackle childhood obesity through enjoyment, movement, nutrition education, and social elements like team play [44]. Providing educational content through them this makes knowledge about nutrition and physical activity more attractive and acceptable to children [45].

2.2. Overview of Proposed Framework

We propose a federated and decentralised framework to tackle childhood obesity leveraging advanced data collection, AI-powered analytics and user engagement Figure 1. The main user interfaces of the architecture are the health application and the serious games suite. The health application serves as a platform for monitoring key health metrics, such as physical activities, screen time, sleep, well-being and providing tailored recommendations. Meanwhile, the serious games leverage gamification techniques to encourage children to adopt healthier behaviors through engaging and educational gameplay. The dashboard helps parents and healthcare providers make evidence-based decisions. The health app and the serious games will also be used to encourage children to adopt healthier lifestyles. A community network and knowledge hub will promote collaboration and disseminate evidence-based resources. The data collected from both the health application and serious games are securely processed and stored within independent Biobank edges that collectively form the Biobank, ensuring privacy and regulatory compliance. Further data harmonisation and pseudonymisation is undertaken to improve data quality and security, and GANs will be used to create synthetic data to support research while maintaining anonymity. In addition, AI tools, such as risk assessment and recommendation engines, are used to provide personalized health insights that complement the dashboard.

2.3. Framework Advantages

Our proposed framework offers several meaningful improvements over existing solutions in key areas. Current data-integration approaches tend to rely on siloed collection methods with limited cross-system compatibility [46,47], whereas our framework introduces a federated Biobank network with harmonized data flows to enhance data richness while respecting privacy concerns. For user engagement, existing systems typically employ single-channel tools with sustainability limitations [48,49], while our multi-channel approach through health apps and serious games aims to support better long-term adherence. Regarding privacy protection, traditional solutions often use centralized storage with conventional anonymization techniques [47], but our framework proposes decentralized edge Biobanks with advanced pseudonymization and GAN-based synthetic data generation to strengthen privacy while maintaining data utility. Finally, current stakeholder collaboration shows limited interaction between healthcare providers, researchers, and families [50,51], which our framework addresses through an integrated dashboard and community knowledge hub designed to facilitate knowledge transfer and collaborative decision making.

3. Description of Framework

3.1. Biobank

The Biobank is a cornerstone for the advancement of research on childhood obesity. It can support improved diagnostic and therapeutic approaches, support the development of prevention strategies and also help to understand the transition from metabolically healthy to unhealthy states.

The proposed Biobank architecture is designed to address the challenges faced by centralized Biobanks, data integration, complexity of ownership, and compliance with privacy regulations, such as GTPR, for example. The system leverages a federated structure that enables decentralized data management while ensuring accessibility, interoperability and security. Each participating site maintains an independent Biobank edge that acts as a local node for data collection and storage. This approach eliminates the need to aggregate sensitive health data, thereby maintaining data sovereignty and ensuring compliance with regional privacy frameworks. The Biobank edge integrates several key components to manage the data lifecycle. Data curation ensures the quality and representativeness of the datasets, while techniques are applied to protect patient privacy by transforming identifiable data into secure, non-traceable formats. The backend of the Biobank edge is built on a high-performance architecture utilizing FastAPI [52] for processing logic and MongoDB [53] for scalable data storage, complemented by a dedicated Query Handler that optimizes data access and retrieval processes.

3.2. Data Management and Preprocessing

The proposed architecture addresses essential data management and preprocessing functionalities to support scalable, secure, and privacy-preserving data processing across distributed environments. Data harmonization resolves inconsistencies that arise from integrating diverse datasets originating from multiple sources [54]. Differences in formats, standards, and semantics are resolved using standardized frameworks, such as CDISC or OMOP, ref. [55] and advanced ontology-driven semantic analysis tools. This alignment ensures that the resulting datasets comply with the FAIR principles (Findable, Accessible, Interoperable, and Reusable), enhancing data usability and interoperability [56]. Sensitive information is transformed to prevent unauthorised re-identification [57]. Cryptographic hash functions, deterministic pseudonymisation and advanced approaches, Merkle trees, secure multi-party computations, are implemented to achieve data pseudonymisation [58]. Communication with the distributed nodes is achieved through RESTful APIs over secure channels, using SSH-based connections to streamline access while maintaining robust security [59].

The architecture implements an authorization framework centered around a node gateway, which acts as a secure intermediary between local Biobank nodes and centralized services. OAuth2-based authentication through a central Keycloak server ensures that only properly authenticated users can access data, with permissions managed through role-based controls. The node gateway provides application-specific and user-level authorization, ensuring users can only access permitted data.

Data generated by applications remains decentralized and is stored directly in Biobank nodes at their respective clinical sites, maintaining data sovereignty by ensuring sensitive information stays within appropriate jurisdictions. Each Biobank node incorporates processing pipelines for harmonization, curation, and pseudonymization before local storage or transmission.

Beyond SSH connections, all data transmission between components utilizes TLS/SSL encryption as detailed in the integration matrix. REST APIs incorporate comprehensive security controls and are documented using OpenAPI specifications. The platform includes a Distributed Log Server and Security Monitoring Service that track all data transfers and API requests for security auditing.

We also use Generative Adversarial Networks (GANs) [60], to generate synthetic datasets that preserve the statistical properties of real data without exposing sensitive information [61]. This reduces dependency on real-world datasets and supporting extensive research and development initiatives.

3.3. AI Tools

The AI tools integrate predictive modeling, behavioral data processing, and personalized recommendation capabilities within the system architecture. The system ensures comprehensive tracking of user engagement metrics [62], while maintaining privacy standards, and analyzes patterns to optimize interventions. In addition, it evaluates individual risk factors related to metabolic conditions through models trained on real-time and historical datasets, enabling the accurate identification of potential health risks. Our goal is to have collected a broad data set of more than 10,000 samples from hospital records and direct user interactions in the final phase of the training with our health and serious games app. By processing user-specific health data and incorporating established guidelines, it provides tailored feedback and actionable suggestions, such as dietary adjustments, physical activity goals, or adherence protocols, that support evidence-based decision making for healthcare professionals and promote healthier behaviors for users, ultimately ensuring that interventions remain targeted, practical, and personalized to individual risk profiles.

For performance evaluation, we employ a combination of metrics including area under the ROC curve (AUC), precision–recall curves, and Cohen’s kappa coefficient, which represent the current state-of-the-art for imbalanced healthcare datasets. We utilize a transfer learning approach, starting with a pre-trained foundation model that is then fine-tuned on our domain-specific data to maximize both efficiency and accuracy. Cross-validation will be implemented using a stratified 10-fold approach to maintain class distribution across folds, with additional temporal validation to account for potential concept drift in longitudinal health data. Data imbalance and ethnic diversity considerations are addressed through strategic sampling techniques and demographic weighting. Specifically, we will take into account variations in disease prevalence across different populations by implementing adaptive boosting for underrepresented groups and ensuring proportional representation across major ethnic categories in our training data. All model output is intended exclusively as a tool for clinicians, who retain full authority for final recommendations and treatment decisions. In addition, all recommendations generated by the AI are validated by healthcare professionals who review sample recommendations and incorporate a feedback loop where user results are tracked to continuously improve the accuracy of the model.

3.4. User Interface and Engagement

The User Interface and Engagement Layer forms a critical front-end component of our proposed architecture, acting as the primary interaction point for children, parents, and healthcare providers.

For children, the platform features an interactive mobile application that promotes healthy living through gamified experiences. This app supports the collection of key behavioral data, including physical activity, dietary habits, and lifestyle metrics, encouraging active participation. Elements such as personalized challenges, task-based rewards, and visual progress tracking foster engagement, transforming health monitoring into an enjoyable and educational experience, motivating children to adopt and sustain healthy habits. For instance, a 9-year-old student plays Food Ninja on the mobile app, earning points by identifying food groups and learning the basics of a balanced diet. The app keeps track of the time spent in the app, the score achieved, and the progress made.

Parents benefit from a dashboard that provides a clear and comprehensive view of their child’s progress and health engagement. This dashboard aggregate behavioral and health data, presenting trends, metrics, and insights in an accessible format. Parents input daily information about their child’s eating habits, the type and duration of physical activity, and have access to historical charts and clinical recommendations. The data is sent to a clinician, who, with the help of AI tools, can provide personalized dietary guidelines and family activities tailored to their preferences. Parents can receive alerts, view personalized recommendations, and gain a deeper understanding of their child’s health status while maintaining secure, role-based access to ensure data privacy and relevance.

Healthcare providers interact with the platform through the dashboard designed to facilitate clinical decision-making and intervention strategies. This dashboard integrates tools like risk-assessment modules, enabling providers also to monitor children’s health data, identify potential risks, and deliver personalized recommendations for healthy living. For example, a pediatrician reviews a patient’s consolidated health data before their annual check-up, noting trends in BMI and activity levels. The dashboard flags potential early warning signs and suggests evidence-based intervention approaches based on the child’s specific engagement patterns and preferences. Visual analytics powered by Matomo Analytics and Prometheus transform complex health metrics into actionable insights, supporting healthcare professionals in making data-driven decisions to improve outcomes.

This layer also serves as a primary interface for data collection and integration, enabling real-time input of behavioral and health metrics through mobile and web channels.

Complementing the User Interface and Engagement Layer, the community network and knowledge hub further enhance knowledge dissemination, and collaboration among stakeholders. The community network acts as a virtual ecosystem where children, parents, and healthcare providers connect, share experiences, and work toward common health goals. Children are encouraged to engage in peer-driven activities, such as community fitness challenges. These activities promote teamwork and healthy habits. Additionally, parents can participate in support groups and educational campaigns to share insights and strategies to better manage their child’s health. For healthcare providers, professional communities enable the exchange of best practices and collaborative approaches to obesity prevention, while feedback channels provide direct communication pathways with families to offer tailored advice and support.

The knowledge hub functions as a centralized repository of resources and tools, empowering all stakeholders with evidence-based information and actionable insights. It provides access to educational materials, guidelines, clinical datasets, and the latest research, which inform effective intervention strategies.

3.5. Serious Games

This section outlines the methodology and development of the serious games proposed that are designed to support the prevention and management of obesity among young populations. Games have been deployed as mobile apps and web-based applications, allowing for easy access.

Table 1. Comparison of serious games for obesity prevention.

Characteristics	Food Ninja	Food Quiz	Food Treasure	Let’s Move
Primary Objective	Food-group identification and categorization	Health and nutrition literacy	Combine physical activity with nutrition education	Establish regular physical activity habits
Target Users	6–12 years	8–16 years	8–14 years	6–14 years
Core Mechanics	Tapping/scrolling items in categories	Multiple-choice questions with aids	AR scanning of hidden items	Guided exercises and dance routines
Learning Focus	Food groups and balanced diet	Meal patterns, nutrition basics, dietary patterns	Nutritional information about specific foods	Exercise techniques and movement patterns
Social Elements	Individual play	Multiplayer option	Parent–child interaction	Family participation
Physical Activity	None	None	Moderate	High
Technology	Basic touchscreen	Basic device	Smartphone with AR capability	Basic device with video playback
Environment	Indoor screen-based	Indoor screen-based	Indoor/outdoor exploration	Indoor or outdoor space for movement
Feedback	Immediate feedback with educational messages	Explanations for incorrect answers	AR information displays	Visual guidance and achievement tracking
Professional input	Nutritional guidelines	Evidence-based questions	Nutritional information	Pediatric consultation for exercises

presents a comparative overview of key characteristics in four serious games designed to prevent obesity. Following this table, we provide an analytical presentation of each game, focusing on its objectives, gameplay mechanics, and educational impact.

3.5.1. Food Ninja

Food Ninja is designed to educate players about healthy lifestyle habits in terms of foods and food groups and meal patterns (Figure 2). The game utilizes a dynamic interface where specific items related to diet and physical activity move across the screen, requiring players to identify and select items from specific categories. The core mechanics involve scrolling items, category-based gameplay, and player interaction through tapping or clicking. Each level focuses on a specific category or subcategory, with points awarded for correct selections. As players progress, they encounter more complex categories and increased scrolling speeds. The game offers two distinct modes of play: Educational and Competitive. In Educational mode, the difficulty increases gradually at a measured pace, allowing players to learn without pressure. If a player makes an error, they can restart the round without losing progress, as there is no game-over mechanic. The Competitive mode features a more challenging experience with faster difficulty progression, where errors result in game over. Scoring in this mode is based on the number of consecutive successful rounds completed.

When players make incorrect selections, they receive immediate feedback displaying the item’s name and proper category, reinforcing learning through error recognition. Educational messages appear after every round, regardless of outcome, to enhance user understanding of the principles of a healthy lifestyle. The game adjusts its difficulty based on the player’s age group and progressively increases in complexity with more specific subcategories and faster gameplay as players improve their skills.

Educational objectives: The primary educational goals of Food Ninja are to improve the ability of players to identify foods within specific food groups, increase understanding of food groups and their role in a balanced diet, encourage critical thinking about food choices through timed decision making, familiarize users with various forms/types of physical activity and its benefits, and provide age-appropriate lifestyle information in an interactive format.

Behavior Change Mechanisms: The time pressure simulates real-world food-choice scenarios, requiring quick decision making similar to grocery shopping or meal-preparation situations. The category-based gameplay directly reinforces nutritional taxonomy knowledge by requiring players to classify foods according to dietary guidelines. Progressive difficulty increases cognitive engagement through spaced repetition, a proven technique for long-term knowledge retention. The immediate feedback addresses misconceptions in real time.

3.5.2. Food Quiz

Food Quiz is an interactive experience with progressively difficult questions across various categories, including meal patterns, physical activity, dietary patterns, nutrition basics, and healthy lifestyle choices (Figure 3). The game’s core mechanics involve answering multiple-choice questions that increase in complexity as players advance. To enhance the gameplay and provide strategic elements, Food Quiz incorporates aids, such as a 50/50 option that eliminates two incorrect answers, or the option to ask a family member for assistance.

The game features two distinct modes: Educational and Competitive.

• In Educational mode, players engage with a predetermined set of theme-specific questions, each presenting four multiple-choice options with one correct answer. When players select incorrect answers, they receive educational messages explaining the correct choice, fostering learning without the pressure of scoring or time constraints.
• The Competitive mode introduces additional challenges through a countdown timer for each question and a sophisticated scoring system that considers both answer accuracy and response speed. Players accumulate points based on consecutive correct answers, with the game ending upon a predefined number of incorrect responses.

Educational objectives: The structure of the game allows for education on important healthy lifestyle concepts and the promotion of health and nutrition literacy (e.g., importance of physical activity and better understanding of healthy eating and dietary patterns), as well as encouragement of family interaction about healthy lifestyle issues. The progressive difficulty and use of aids help maintain player engagement while providing a sense of accomplishment as they advance through the game.

Behavioral Change Mechanisms: The multiple-choice question format directly transfers factual nutrition knowledge while the progressive difficulty structure creates a scaffolded learning experience that builds conceptual understanding incrementally. The family assistance option deliberately introduces social support as a behavioral change technique, normalizing health discussions within the family unit.

3.5.3. Food Treasure

Food Treasure is an Augmented Reality (AR) game that combines physical activity, family engagement, and nutrition education in a unique and interactive format (Figure 4). This treasure hunt-style game is designed to combat physical inactivity while providing educational content about food and nutrition. The core mechanic is that a parent or teacher hides either a QR code or an actual food item in a designated area. Children must search the area in order to scan for the hidden item. Once found, they can use their mobile phones to interact with the found item. Through an AR application on their mobile device, players receive nutritional information about the food and, in the case of QR codes, see a 3D projection of the food in their real-world environment.

Educational objectives: By encouraging active learning about nutrition, reducing sedentary time, and promoting family interaction, Food Treasure addresses several challenges that modern families face. Its adaptable nature makes it suitable for a variety of settings, from homes to schools, while the use of AR technology creates an engaging and immersive educational experience.

Behavioral Change Mechanisms: The game seamlessly integrates physical activity with learning. The AR scanning interaction creates a powerful connection between abstract nutritional concepts and tangible food items, bridging the knowledge–behavior gap through experiential learning. The parent–child collaborative design intentionally positions caregivers as co-learners. While the “hidden item” game structure models food exploration and encourages children to approach unfamiliar healthy foods with curiosity.

3.5.4. Let’s Move

Let’s Move is a physical activity game designed to establish consistent exercise habits in children (Figure 5). The game features two primary modes: a personalized exercise mode and an interactive dance mode, both designed to promote family engagement and sustained physical activity.

The personalized exercise mode employs a progressive difficulty system that gradually increases activity duration, utilizing exercises that are individually tailored to each child’s needs and capabilities through professional pediatric consultation. The core mechanics revolve around daily exercise routines that are carefully calibrated to meet the recommended 60 min of moderate aerobic activity for children. The game implements an adaptive progression system in which the duration of exercise increases by 10% over time, allowing children to build endurance and confidence gradually.

Dance mode offers a family-friendly physical activity, where children and family members can learn and perform choreographed movements together. This mode promotes social interaction, coordination, and rhythmic awareness while maintaining benefits of physical activity. Dance routines are gradually introduced, allowing families to build their repertoire of movements while enjoying quality time together. The collaborative nature of this mode helps strengthen family bonds while encouraging regular physical activity as a shared experience. This structured approach ensures that players can work safely and effectively to meet the daily physical activity recommendations while developing fundamental movement skills.

Objectives: The primary goals of Let’s Move are to establish consistent physical activity habits, improve cardiovascular fitness through moderate aerobic exercise, develop fundamental movement patterns through structured routines, and combat childhood obesity through regular physical activity participation. Additional objectives include fostering family bonding through shared physical activities, developing rhythmic awareness and coordination through dance, and creating sustainable, enjoyable exercise habits that extend beyond individual practice to include family participation. The progressive nature of the game, combined with professional pediatric oversight, helps children build tolerance to exercise while maintaining motivation through achievable daily goals tailored to their individual needs.

Behavioral Change Mechanisms: The personalized difficulty progression mechanic directly applies the exercise science principle of progressive overload, gradually increasing capacity without triggering exercise aversion. The 10% increment system is specifically calibrated to align with pediatric exercise guidelines while remaining below the threshold that typically triggers dropout. The structured daily routine establishes consistency. By incorporating familiar dance movements rather than formal exercise terminology, the game reframes physical activity as play rather than obligation, addressing a key psychological barrier to sustained activity. The synchronized family movements create social reinforcement through shared experience.

3.6. Framework Adaptability Across Diverse Contexts

Our framework is designed with flexibility to accommodate diverse implementation contexts. The decentralized architecture supports adaptability across varying resource settings, cultural backgrounds, and technological infrastructure. In resource-limited environments, the framework can be simplified by reducing the complexity of Biobank edges while maintaining core functionality. This modular approach allows communities and healthcare systems to adopt specific components based on their available resources and needs. Implementation can range from basic data collection in areas with minimal infrastructure to comprehensive biospecimen analysis in well-resourced settings. Navigating challenges across jurisdictions and institutions is critical for our framework. This involves developing modular content that allows institutions to customize interventions while maintaining core evidence-based elements, establishing partnerships with local health authorities, schools, and community organizations to address regulatory requirements. Efforts are made to ensure that the UI/UX design accommodates children with different abilities and learning needs. Color combinations that are colorblind accessible, auditory cues, and simplified navigation make the tools more accessible to a broader audience. To address digital divide challenges, key components of the health application and serious games include offline functionality. This allows users to interact with the system and synchronize data when connectivity becomes available. In areas where digital access remains limited, the framework incorporates alternative engagement pathways through existing community structures such as schools, community centers, and religious institutions. Cultural adaptability is embedded throughout the system. All user-facing components—including health app, serious games, community network and educational resources in the knowledge hub—can be translated and culturally adapted to local contexts. Educational messages can be customized to reflect local cultural settings. This, for example, includes tailoring dietary recommendations to feature locally available foods. The knowledge hub can be populated with regionally specific health information, while the AI recommendation system can be fine tuned on local datasets to provide contextually appropriate suggestions. By combining technical flexibility with cultural adaptability, the framework can effectively address childhood obesity across diverse global contexts while respecting local constraints and cultural differences.

3.7. Regulatory Compliance and Ethical Considerations

Our framework implements GDPR compliance through technical and procedural safeguards. The decentralized architecture aligns with data minimization by processing data locally, reducing cross-border transfers. Purpose limitation controls restrict data usage to disclosed research aims, while audit logs capture all processing activities per Article 30. For children’s health data, we have established enhanced safeguards including encrypted pseudonymization that separates identifiers from health data, creating barriers to re-identification. DPIAs are conducted before new processing, with focus on children’s vulnerabilities. Parents and age-appropriate children can access a portal to view data usage and exercise their rights to access, correction, and erasure. International collaborations use standard contractual clauses with additional child-protection provisions. Our retention policy balances research needs with data minimization, applying distinct schedules based on scientific necessity. Technical controls enforce retention limits through scheduled pseudonymization and data transformation. As data age, identifying elements are progressively removed or encrypted, creating a privacy gradient. Retention schedules include developmental milestones that trigger reviews. A governance committee with parent representatives periodically reassesses retention practices to ensure alignment with research objectives, ethical standards, and children’s best interests. The system maintains documentation of decisions and their justification for regulatory review.

Verifiable parental consent is obtained before any processing occurs. For data governance, age-based access controls are implemented with different policies for different age groups of minors, restricting the scope of accessible information based on need. The decentralized nature of our system further supports regulatory compliance by enabling region-specific policy enforcement at each biobank node, automatically applying the appropriate regulatory framework based on the location of the data. All pediatric data transactions are specifically flagged in our security monitoring system, creating dedicated audit trails that document every instance of access or processing involving children’s health information. A dedicated opt-out process allows participants to withdraw from studies and have their data removed. The platform creates anonymous user IDs for participants, which are synchronized with the authentication server to maintain privacy while enabling necessary functionality.

Biospecimen sampling procedures are designed to be minimally invasive, with age-appropriate collection protocols that prioritize children’s comfort and psychological well-being. Our ethical framework mandates periodic re-consent at key developmental transitions (e.g., entering adolescence) to ensure continued voluntary participation as children mature and develop greater decision-making capacity.

Our framework addresses ethical concerns in health gamification for children. We avoid competitive mechanics that might create unhealthy fixations on body image, focusing instead on positive behavioral achievements. Game messaging is crafted to prevent anxiety or shame around eating behaviors. The system limits engagement mechanisms to prevent addiction, with mandatory breaks and age-appropriate session limits. Clear boundaries exist between educational content and entertainment, ensuring scientific accuracy in health messaging. Child psychologists review each game for emotional impact and developmental appropriateness. To prevent data exploitation, the games collect only necessary gameplay data for measuring intervention effectiveness, avoiding excessive tracking. We maintain strict separation between research and commercial gaming ecosystems to protect vulnerable children. Our approach develops intrinsic motivation rather than dependency on external rewards, ensuring health behaviors stem from genuine understanding.

3.8. Framework Evaluation

Following design science principles [63], we evaluate our proposed architecture against traditional approaches to childhood obesity management. This evaluation focuses on how well our federated framework addresses the explicated problem and fulfills the defined requirements compared to existing solutions.

Table 2. Architecture evaluation according to design science principles.

Evaluation Criterion	Our Federated Architecture	Traditional Centralized Approaches	Evidence of Improvement
Data Privacy and Security	Decentralized Biobank edges with local data storage and pseudonymization techniques	Centralized data repositories with conventional anonymization	Reduced risk of large-scale breaches; GDPR compliance through data sovereignty; advanced pseudonymization; and synthetic data generation
Data Integration	Harmonized data flows across distributed nodes with standardized frameworks (CDISC/OMOP)	Siloed data collection with limited cross-system compatibility	Enhanced data richness while maintaining privacy; enables comprehensive analysis across multiple sources
Scalability	Modular components that can be implemented independently; MongoDB for scalable storage	Often requires complete system implementation; limited by central server capacity	Institutions can adopt specific components based on resources; distributes computational load across nodes
User Engagement	Multi-channel approach through health app and serious games suite	Single-channel tools with limited engagement strategies	Better long-term adherence through gamification and personalization; targets multiple behavioral factors simultaneously
Clinical Decision Support	AI-powered risk assessment and recommendation engines with healthcare professional oversight	Manual interpretation of fragmented data sources	Evidence-based decision making with integrated dashboard; personalized intervention recommendations
Stakeholder Collaboration	Integrated dashboard and community knowledge hub	Limited interaction between providers, researchers, and families	Facilitates knowledge transfer and collaborative decision-making; creates feedback loops between stakeholders
Adaptability	Flexible implementation options for varying resource settings and cultural backgrounds	Often requires standardized implementation	Components can be adopted based on local constraints; culturally adaptable content and recommendations
Technical Implementation	RESTful APIs with standardized documentation; modular architecture	Proprietary interfaces, monolithic systems	Easier integration with existing healthcare systems; standards-based approach reduces implementation barriers
Ethical Considerations	Verifiable parental consent process; age-based access controls; opt-out mechanisms	Often limited privacy controls	Enhanced protection for minors’ data; clear governance structure for sensitive information

presents a systematic comparison across key evaluation criteria.

The evaluation demonstrates several key advantages of our proposed architecture. First, the decentralized Biobank edges provide enhanced data privacy while still enabling comprehensive analysis through harmonized data flows. This directly addresses limitations of centralized approaches [46,47], that struggle to balance data utility with privacy protection. Second, our multi-channel user engagement strategy leverages both health applications and serious games to create sustainable behavior change, overcoming the engagement limitations observed in single-channel interventions [48,49]. Third, the framework facilitates improved stakeholder collaboration through integrated dashboards and knowledge sharing platforms, enhancing communication between healthcare providers, researchers, and families compared to traditional siloed approaches [50,51]. Finally, the modular nature of our architecture enables contextual adaptation across diverse implementation settings, allowing components to be adopted based on local resources, cultural considerations, and technical capabilities.

Mapping Architectural Components to Requirements

Our architecture fulfills critical requirements through targeted design choices (

Table 3. Architecture components fulfilling obesity management requirements.

Requirement	Architectural Component	Enabling Features
Data Privacy Protection	Federated Biobank Edges	Decentralized storage keeping data at originating sites Advanced pseudonymization processing GAN-based synthetic data generation
Comprehensive Data Integration	Data Harmonization Layer	Standardized frameworks (CDISC/OMOP) Ontology-driven semantic analysis RESTful APIs over secure channels
Sustainable User Engagement	Multi-channel User Interface	Mobile health application Serious games suite Age-appropriate gamification elements
Evidence-based Intervention	AI Tools and Knowledge Hub	Risk-assessment models Personalized recommendation engine Knowledge repository for healthcare professionals
Multi-stakeholder Collaboration	Community Network and Dashboard	Role-based interfaces for children, parents, and clinicians Shared visualization of progress Collaborative decision-making tools

). The decentralized Biobank structure with independent edges directly addresses privacy concerns by ensuring sensitive health data remains within appropriate jurisdictions while still enabling secure analytics. The node gateway component functions as a secure intermediary that enables controlled access to data through application-specific and user-level authorization protocols, ensuring users can only access permitted data. To facilitate comprehensive data collection while maintaining user engagement, we implemented a dual approach through the health application and serious games suite. These components work synergistically—the health application captures structured health metrics while the serious games provide an engaging context for behavior modification and additional data collection. The dashboard’s visualization capabilities transform complex health metrics into intuitive insights, allowing parents and healthcare providers to identify trends and make evidence-based decisions.

4. System Interaction Workflows

This section presents sequence diagrams depicting the critical data flows and user interactions within our Biobank architecture.

The sequence diagram in Figure 6 illustrates data analytics architecture in which multiple applications share anonymous user IDs within a unified system. The five applications (ActiveHealthApp and four serious games) are connected to a centralized Matomo Analytics platform. The ActiveHealthApp uses server-side data capture to transmit information to the analytics platform, while the four serious games communicate via REST API. This enables consolidated tracking and analysis of user behavior across all applications. The system is designed to collect and analyze user engagement data in a privacy-conscious manner.

User interactions with the knowledge hub are represented in Figure 7, which shows the communication flow between users, the ActiveHealth App, and the knowledge hub. When a user starts the app, one of two paths occurs based on authentication status: authenticated users navigate to a home page and choose a website from navigation options, while unauthenticated users view a welcome page with a slider from which they select a website. In both scenarios, after the user makes their selection, the ActiveHealth App launches the chosen website by communicating with the knowledge hub, which then displays the relevant website content back to the user through the app. This streamlined process ensures all users can access appropriate content regardless of their authentication status, with the knowledge hub serving as the central content repository that delivers information back through the ActiveHealth App interface.

As shown in Figure 8, the serious game system follows a structured interaction pathway beginning with user authentication. The sequence starts with user authentication through the AUTH service to access the dashboard. Once authenticated, users can interact with the game through start and resume actions handled by the SERG (serious game) component. During gameplay, user interaction data are temporarily stored on the user’s device. Upon game completion, after processing, the collected data are transmitted to the Biobank edge through the node gateway (NGW) service. When the game data reach the node bundle, it undergoes processing through harmonization, curation, and pseudonymization components before being securely stored. For AI model processing, authorized healthcare professionals can access these data through the dashboard, using secured REST APIs. This allows AI components to analyze the game data when triggered by an authorized user through the dashboard interface. The entire data flow is protected by OAuth2 authentication via the central Keycloak server, ensuring that only properly authenticated and authorized users can access sensitive information throughout the system.

5. Conclusions and Future Work

The challenge of childhood obesity requires innovative approaches that take advantage of technology while maintaining privacy, security, and engagement. Our proposed federated and decentralised framework represents a comprehensive solution to this complex public health issue. By integrating independent Biobank edges for localized data collection with edges in schools and communities, we create a robust data ecosystem that respects privacy concerns through careful data harmonization, pseudonymisation, and synthetic data generation via GANs. The strength of the framework lies in its multifaceted approach, combining AI-powered analytics for personalized risk assessment with user-friendly dashboards that enable evidence-based decision making by parents and healthcare providers. The incorporation of serious games and health applications engages children directly, encouraging healthier lifestyle choices through interactive experiences that target both protective factors and risk factors. Furthermore, the community network and knowledge hub foster collaboration and knowledge sharing among stakeholders. Economic sustainability remains central to our approach, with cost assessment tools ensuring the viability of implementation across diverse settings.

As a critical next phase in our research, we plan to subject the entire system to a comprehensive ethical review through an Institutional Review Board (IRB). This formal evaluation will ensure that all aspects of data collection, storage, processing, and usage comply with ethical standards and regulatory requirements.

We plan to implement a comprehensive validation strategy through workshops and pilots. As a first step, we will conduct a series of workshops with at least 200 patients and students to refine the games based on real-world needs and contextual factors. These workshops will inform the final protocol design by incorporating cultural considerations and research-related factors. This participatory approach will ensure that the proposed architecture meets real user needs before formal testing.

The next step will be to implement a mixed-methods validation study over one school year with a pre-test/post-test experimental design comparing control and intervention groups. The intervention will span 6 months and include four joint parent–child classroom workshops focusing on healthy eating and physical activity to promote preventive behavior. Key outcome measures will include changes in nutrition knowledge scores, daily physical activity levels measured via validated instruments, food-choice behaviors in controlled settings, and user engagement metrics. The pilots will be conducted in collaboration with established obesity care centers where multidisciplinary teams including child psychiatrists, psychologists, endocrinologists, diabetologists, surgeons, and nutritionists will support the implementation and assessment. These pilots aim to demonstrate the impact of the risk assessment framework on clinical routine and guidelines, the effectiveness of the recommendation system on patient empowerment and engagement, and the potential for understanding how multiple factors (genetic, epigenetic, environmental, socioeconomic, and lifestyle) interact in the development of childhood obesity.