As the field of sports science continues to progress, the integration of cutting-edge technologies has become increasingly crucial to strengthen the long-term athlete development processes. One such transformative technology is AI, which has the potential to revolutionize the approach to talent (Faqihi and Miah, 2023). AI has already demonstrated its adaptability across various domains, including education, healthcare and business (Zhang et al., 2022). In the field of sports, AI-powered methods could be applied to analyze massive raw data, identify patterns, and make informed decisions that can significantly improve the accuracy and efficiency of Talent Identification and Development (TID) (Faqihi and Miah, 2023; Cossich et al., 2023). By using AI-driven algorithms, sports practitioners could now analyze a multitude of factors, such as an athlete’s physical attributes, technical, tactical and psychological to create comprehensive player profiles (Fu, 2020). This holistic approach to talent identification can help professionals uncover hidden gems, ensuring that the most promising athletes are identified and developed.

Adolescence represents a critical developmental window in youth sports, marked by substantial physiological and psychological changes (Güllich et al., 2022; Leite et al., 2021; Young et al., 2010). An accurate assessment of maturation processes is central to refining training routines and selection strategies, with the dual aim of enhancing performance (Cobley et al., 2009) and reducing health-related risks (Towlson et al., 2021). The recognition that variations in maturation timing exert significant influence on athletic performance, selection outcomes, training efficacy (Meylan et al., 2010) and injury proneness highlights the imperative for nuanced understanding of child athletes development (Arede, 2021). Despite this importance, the effectiveness of training and selection practices ultimately depends on the validity and reliability of the methods used to estimate GAM.

Conventional techniques for determining GAM status, such as radiographic methods to determine skeletal age (Chumela et al., 1989; Morris, 2003) and measuring sexual maturity based on Tanner stages (Malina et al., 2005; Rai et al., 2014), have limitations despite being useful. Radiography entails significant financial expenses, including the cost of X-ray imaging equipment, maintenance, and the need for trained radiologists to interpret the results. For instance, the average cost of a single X-ray scan ranges from $100 to $500 depending on the healthcare facility and region (Chumela et al., 1989; Morris, 2003). These costs can accumulate substantially in sports contexts, especially when repeated imaging is required for longitudinal monitoring.

Assessments of sexual maturity, in contrast, may be constrained by subjectivity, reliance on self-reported information, and privacy concerns when evaluating youth athletes (Fragoso et al., 2014; Matsudo and Matsudo, 1994). Although sonography provides a non-invasive alternative for estimating skeletal maturation (Guglielmi et al., 2009; Utczás et al., 2017), still it has limitations due to the array in the rate of bone development and occasional mistakes in identifying different phases of maturity.

While those traditional methods for assessing maturity have contributed substantially to GAM assessment, their limitations still necessitate to explore new approaches that are non-invasive, inventive, and cost-effective. The progress in statistical modeling, ML, and digital health technologies is revolutionizing the field of GAM research (Miotto et al., 2018). These advancements offer scientifically rigorous and practical methods that may be applied in several areas. As these technologies continue to evolve, they hold considerable potential for enhancing both the precision and scalability of GAM monitoring in young populations (Alber et al., 2019; Marsch, 2020).

The demand for non-intrusive and field-applicable methods of GAM monitoring has led to the development of several linear and nonlinear growth modeling techniques (Freeman et al., 1995; Grimm et al., 2011; Ostojic et al., 2014), allow for security data collection in real-world. Somatic methods derived from anthropometry have gained wide acceptance in soccer academies and other sports organizations for maturation indexing, determining the temporal distance from peak height velocity (PHV), and forecasting adult height, backed by standardized protocols from national regulatory bodies (Cumming et al., 2018; Fransen et al., 2021; Salter et al., 2022). Those methods still need the involvement of health experts or sports practitioners to control the estimation processing and explain the use of the GAM information to the players and guardians (coaches and parents).

Recent developments in mobile technologies have incorporated advanced AI methods such as CV, ML, and deep learning (DL) in personal devices. CV consists of a set of complex image recognition processes capable of detecting and classifying features from multimodal inputs. Combined with ML algorithms, CV can recognize human postures and joint points in images and subsequently process this information using DL algorithms to generate deeper insights (Khanal et al., 2022).

In parallel, several professional institutions have developed online tools for growth and maturity assessment, including the UK-WHO Growth Charts: 2–18 Years by the Royal College of Pediatrics and Child Health (RCPCH), the Maturity Status Calculator provided by the Office of Sport, NSW Government in Australia, and the Bioband Your Child tool from the University of Bath. These tools support the evaluation of growth patterns, biological maturity, and developmental stages in children and adolescents, offering valuable insights for healthcare professionals, educators, and re-searchers. But they are only with easy description of results (the biology age and growth status), without more deep info and guidance for players and coaches to use that. Also, the online calculator cannot automatically track the consistent GAM information and that is useful and good for longitude monitoring sport players. Other growth and maturity (GAM) monitoring tools like BoneXpert (which utilizes dorsopalmar radiographs) and BAUSport (which relies on ultrasound assessments) require professional health assistance, adding to their cost (ranging from $100 to $500 for radiographs/ultrasound assessments and software licenses) and complexity.

By addressing the limitations of traditional methods and offering scalable solutions for large-scale use (Baars et al., 2022; Bernaciková et al., 2022), AI technologies have the potential to transform growth monitoring (Guigas, 2017; Straun et al., 2024). Nonetheless, the transition to AI-based systems must proceed with caution, ensuring that their adoption is guided by rigorous validation to establish reliability and safety.

In response to these needs, our team developed an app with AI methods used in GAM of youth. This paper aims to describe a mobile app named Maturo that is designed to monitor the GAM of youth, that it could provide youth athletes and their guardians with awareness about the GAM information using an easy-to-use and low-cost tool that can be a good approach for helping to strength the long-term athlete development processes. Maturo integrates automated data capture, longitudinal growth tracking, advanced analytical models, and a user-friendly graphical interface, which could offer a practical and scalable solution for real-world youth sport settings.

The development of the Maturo app was the result of a collaborative initiative involving basketball and football coaches from Vila Real (Portugal) and researchers from the School of Science and Technology (ECT) and the Research Center in Sport, Health, and Human Development (CIDESD) at the Universidade de Trás-os-Montes e Alto Douro (UTAD). The current demo version of the app restricts access exclusively to healthcare professionals. Coaches and parents can only review data in collaboration with these professionals. Future iterations will implement a robust role-based access system, assigning tailored permissions to athletes, coaches, and healthcare providers.

Maturo has been designed to accommodate a broad range of users, from young athletes to adults, and includes an educational portal that supports coaches, guardians, and teenagers in understanding GAM information. The application is engineered as a mobile application, utilizing the flexibility of Dart to provide a distinct experience on Android platforms, enabling smooth integration of CV and ML model components. This strategic choice not only improves user accessibility but also could have a fast development cycle and maintenance operations.

The user interface follows contemporary principles from Material Design (Version 3, Google, 2024), to provide responsive layouts, straightforward navigation, and captivating animations. The app employs a simplified navigation system with clearly labeled buttons and intuitive workflows, reducing cognitive load for users with limited technical expertise (e.g., younger athletes, parents unfamiliar with technology). And Material Design’s framework emphasizes contrast, legible typography, and accessible color schemes, which are particularly beneficial for users with visual impairments or color blindness. These features aim to enhance usability and encourage active user participation (Ahmadi and Kong, 2012).

Maturo applies YOLO (You Only Look Once) v7 algorithms to enhance the precision and efficiency of its maturation assessments, and all algorithm and the following information processing are run locally on the device. YOLO v7 is notable for its architecture based on F-CNN (Fully Connected Neural Network), which enables faster and more accurate deductions (Wang et al., 2023). The image processing pipeline begins with preprocessing steps such as noise reduction, image normalization, and segmentation to extract critical features from anthropometric images. This ensures that data input quality is consistent and optimized for subsequent analysis. The DL component, built on convolutional neural networks (CNN), processes these extracted features to identify patterns indicative of biological maturation (based on the Khamis-Roche method). All application data is stored locally on the user’s device, without reliance on external servers. This approach eliminates concerns related to server-side data processing and ensures that all personal information remains under the user’s control.

When an athlete uploads an image, the YOLO model identifies key growth markers, such as standing height, sitting height, and limb ratios, through bounding box detection. These detected features are then processed to calculate critical metrics, which feed into the app’s biological age estimation model. The pre-trained YOLO model used in this app was fine-tuned with a custom dataset of 1775 manually tagged images of youth athletes, ensuring its relevance to GAM assessment. The app processes images locally on the user’s device, ensuring both real-time feedback and data security. By using multiple layers of feature extraction and classification, the system can estimate biological age and predict growth trajectories with high accuracy.

To ensure adaptability across diverse populations, the algorithms incorporate transfer learning, enabling the model to fine-tune its predictions based on new data while leveraging knowledge from the original training set. This combination of advanced image processing and DL enables Maturo to automate complex assessments, minimizing the need for manual intervention while maintaining consistency and accuracy.

This study addresses the development and implementation of the Maturo method, highlighting its comparative advantages in real-time GAM tracking. Maturo is an innovative program that uses AI techniques (CV and ML) to automate the assessment and training design in youth sports. By including several data and recognizing individual differences, these models provide a strong basis for assessing maturation using non-invasive methods. In addition, the rise of ML has brought up new techniques for predicting growth and categorizing maturity (Dallora et al., 2019).

Monitoring GAM has become an essential component of contemporary youth sport systems, which helps the coaches to offer tailored programs that align with players’ physical conditions, tactical requirements, and stages of development. This approach maximizes player development while minimizing the risk of injuries. This is particularly relevant for youth players, when rapid growth spurts during adolescence are believed to significantly increase the risk of injuries among young athletes (Johnson et al., 2022). Therefore, implementing scientifically monitoring during this critical period becomes essential. A key takeaway from Salter’s research is the emphasis on the need for accurate biological maturity estimation in the context of soccer academies, particularly for improving injury prevention strategies and tailoring developmental programs for young athletes. However, the consistent application of such practices is often constrained by limited resources, time, and staff expertise (Salter et al., 2022).

Several professional institutions have introduced online calculators for estimating growth and maturity, yet many of these tools require institutional licenses or specialized access, making them inaccessible to parents, coaches, or grassroots sport programs. This barrier reduces their utility in community sport settings where affordability, simplicity, and frequent monitoring are essential.

Maturo differs fundamentally from these systems by functioning solely through a smartphone application, eliminating the need for costly external devices. This significant reduction in equipment costs makes Maturo a far more cost-effective solution. Furthermore, the ease of use associated with an APP means that regular, even weekly, assessments are feasible without requiring extensive setup or specialized training. This affordability and simplicity are crucial for enabling consistent monitoring and timely interventions, ensuring athletes’ developmental needs are met promptly and efficiently.

Beyond sports context, Maturo holds considerable potential for pediatric growth monitoring, offering critical insights for medical professionals and parents alike. Maturo may enable early interventions by properly identifying children who are developing either earlier or later than expected standards, based on precise predictions of development speed. Early treatments are essential for addressing any health or developmental problems, enabling customized assistance and maximizing growth trajectories.

The development of Maturo is grounded in established theoretical perspectives on youth growth, biological maturation, and technology-supported assessment. First, the estimation of biological maturation draws on widely adopted growth and maturation models, particularly the maturity offset and predicted adult height frameworks proposed by Mirwald et al. (2002) and later refined by Malina et al. (2005). These models highlight the nonlinear and individually variable nature of adolescent growth, emphasizing the importance of accessible and repeatable assessment tools for youth sport and health settings.

From a developmental perspective, monitoring biological maturation is central to long-term athlete development pathways. The Long-Term Athlete Development (LTAD) framework (Balyi, 2004) and the Developmental Model of Sport Participation (Côté et al., 2003) both underline the need to align training with biological, rather than chronological, maturation. These models provide a conceptual rationale for tools that can inform coaches, parents, and young athletes about maturation timing in a non-invasive and user-friendly manner.

Given that digital tools increasingly support youth health management, the adoption and use of Maturo can also be situated within technology acceptance and human–computer interaction theories. Constructs from the Technology Acceptance Model (Davis, 1989) and related digital health engagement literature emphasize perceived usefulness, ease of use, and trust as key determinants for the uptake of health monitoring applications by adolescents and their caregivers. These frameworks explain why intuitive design, device accessibility, and transparent data presentation are necessary components of a maturation assessment tool.

This study was conducted in accordance with established ethical standards for research involving minors and was approved by the institutional ethics committee. All participants and their legal guardians provided informed consent prior to data collection. To ensure participant autonomy and comprehension, procedures were explained in age-appropriate language, and assent was obtained from adolescents in addition to parental consent.

Given that the present work represents an initial exploratory validation of the automated maturation assessment approach, several precautions were taken to safeguard the participants’ rights and privacy. All data collected through the study were anonymized prior to analysis, and no personally identifiable information was stored within the system or used for any stage of model development. Data collection and interpretation were supervised by a multidisciplinary team including coaches, sport science specialists, and computer science researchers to minimize risks associated with measurement, interpretation, and data handling.

Recognizing the importance of broader ethical governance in digital health tools, especially those involving minors, we acknowledge that additional expertise is essential for future development. To address the concerns surrounding autonomy, informed consent, data protection, and responsible interpretation, future versions of the application will be guided by an interdisciplinary advisory board consisting of pediatricians, data protection specialists, and ethics scholars. This advisory structure will support the refinement of data governance practices, ensure compliance with evolving international regulations, and strengthen user protection as the system transitions from research prototype to real-world implementation.

The present manuscript represents the first stage of a broader project on GAM monitoring and has already been piloted with more than 200 youth athletes to evaluate the usability of the app in sport environments. Given that PHV age is a second-order temporal feature derived from individual growth-velocity curves, its estimation from cross-sectional anthropometric data inherently results in moderate reliability. This structural limitation has been consistently documented in previous maturation research, where accurate identification of the maximum height-velocity inflection point typically requires longitudinal measurements.

In addition to the maturation prediction approaches most commonly cited in the international literature, it is important to acknowledge that alternative models and population-specific investigations have been developed in different research contexts. Notably, studies conducted in Brazilian youth populations have contributed valuable insights into biological maturation assessment and pubertal prediction, emphasizing the influence of demographic, socioeconomic, and developmental variability on growth trajectories (Mm et al., 2015; de Almeida-Neto et al., 2025; Cabral et al., 2013). These investigations highlight that maturation-related indicators may exhibit population-specific characteristics and that predictive accuracy can vary across cohorts with distinct biological and environmental profiles.

Recognizing this body of work reinforces the need for caution when generalizing maturation estimates across diverse youth populations. In this context, the Maturo application is not intended to replace established methods or to assume universal predictive validity, but rather to provide a practical and scalable tool that integrates established growth frameworks while remaining adaptable to future population-specific calibration. Future validation studies will explicitly explore cross-cultural performance and model generalizability by incorporating datasets from different regions and developmental contexts.

In this exploratory prototype study, correlation analyses were used deliberately to quantify measurement association rather than causal dependency. Because Maturo is still in an early development phase and the sample size is limited, dependency modeling or regression analysis was not conducted to avoid over-interpretation of preliminary results. Future studies with larger multi-timepoint datasets and longitudinal follow-up will incorporate predictive modeling, advanced calibration approaches, and more robust curve-reconstruction techniques. A separate validation study comparing Maturo estimations with expert assessments has been published elsewhere and provides a more comprehensive analysis of measurement accuracy.

The present manuscript is explicitly positioned as a developmental description of the design rationale, functional architecture, and applied feasibility of the Maturo smartphone application for growth and maturation (GAM) monitoring in youth populations. Although selected quantitative indicators, such as intraclass correlation coefficients, technical error of measurement, and cross-tabulation of maturity classifications, are reported, these analyses are included solely to demonstrate system operability and practical agreement, rather than to provide comprehensive algorithmic validation. A separate, previously published study has provided a detailed outcome-level validation of Maturo-derived maturation estimates against expert-based assessments, focusing on measurement agreement and applied reliability in youth sport settings (Shang et al., 2025).

A further limitation relates to the current prototype’s reliance on a small development team, which constrains the speed and scope of feature expansion and technical refinement. As a low-cost application, Maturo does not yet provide all functionalities seen in larger, resource-intensive commercial systems, and continued development will require sustained collaboration and external funding. These limitations highlight the necessity of future studies assessing long-term usability, ecological validity, and effectiveness across diverse real-world training and health-monitoring settings.

The Maturo app could provide automated monitoring of key physical characteristics, including biological age, growth timing, and peak height velocity. As an early-stage prototype, the current version demonstrates the feasibility of integrating computer-vision techniques and anthropometric prediction models into a practical tool for youth sport and health contexts. While preliminary findings suggest that Maturo may offer a low-cost and user-friendly alternative to traditional assessment methods, its potential applicability in underserved or resource-limited regions remains hypothetical, as this study did not directly evaluate such settings. Therefore, further research, including multi-site testing, cross-cultural validation, and long-term real-world deployment studies, is required to determine the app’s effectiveness, usability, and broader accessibility.