سامانه کنگره های دانشگاه علوم پزشکی و خدمات بهداشتی و درمانی آذربایجان غربی

صفحه اصلی / بیستمین همایش سالیانه بیماری های شایع گوارش و کبد کودکان ایران و دومین همایش بین المللی چاقی کودکان

AI-Driven Algorithms for Childhood Obesity Risk Prediction: Integrating Clinical Data and Lifestyle Factors

نویسندگان :

Ladan Soltanzadeh¹ Ali Mirzaee AghGonbad² Shanli Mirzaee³

1- دانشگاه ارومیه 2- دانشگاه تبریز 3- ارومیه

کلمات کلیدی :

Artificial intelligence،Machine learning،Pediatric obesity،Childhood obesity،Clinical data،Lifestyle factors،Risk prediction،Early detection،Prevention strategies

چکیده :

Background and Aim: Childhood obesity is a critical global health challenge linked to cardiovascular diseases, type 2 diabetes, and metabolic syndrome, with long-term impacts persisting into adulthood. Traditional statistical models often fail to capture the complex interplay of genetic, clinical, and lifestyle factors. AI-driven algorithms offer improved risk prediction by integrating high-dimensional clinical data and lifestyle variables, enabling early, personalized interventions. This study aims to develop and evaluate AI models for early obesity risk prediction to enhance prevention strategies. Methods: A systematic review (2020–2025) analyzed 38 studies from PubMed, Scopus, and Google Scholar using keywords like "artificial intelligence," "childhood obesity," and "lifestyle factors." Inclusion criteria focused on original research, systematic reviews, and meta-analyses involving children (0–18 years). Data extraction included AI methodologies (e.g., neural networks, random forests), clinical variables (BMI percentile, family history), and lifestyle factors (diet, physical activity). Quality assessment utilized PRISMA and QUADAS-2 tools. Results: AI models achieved 78–89% accuracy and AUC values of 0.85–0.92, outperforming traditional methods (AUC: 0.72–0.79). Key predictors included BMI percentile (92% of studies), parental obesity (68%), and sedentary behavior (improved accuracy by 12–15%). However, only 21% of studies validated models across diverse populations, and complex algorithms often lacked interpretability frameworks like SHAP or LIME. Conclusion: AI-driven algorithms demonstrate strong predictive performance for childhood obesity but face challenges in generalizability and clinical interpretability. Future research must prioritize equitable data collection and explainable AI techniques to enable actionable, personalized interventions and reduce long-term health burdens.