Paul Yi, MD |St. Jude Children’s Research Hospital; Pritam Mukherjee, PhD; Stephanie Dixon, MD, MPH
Competition Sponsor: National Academy of Medicine
Awardee Year: 2025
Accurately assessing biological aging is essential for identifying individuals at risk for accelerated aging, enabling timely intervention. While epigenetic clocks provide an objective measure of biological age, their high cost and limited accessibility hinder widespread clinical adoption. We propose an AI-driven approach to extract aging biomarkers from routine medical imaging (CT, DEXA) and laboratory tests (e.g., CBC)—datasets already collected in clinical practice but underutilized for aging assessments. Using a unique cohort of over 6000 childhood cancer survivors, a population at high risk for accelerated aging, we will train AI models to predict biological age based on longitudinal imaging and lab data. To ensure generalizability, we will validate these models on both an independent hold-out set of cancer survivors and an external test set of age and sex-matched non-cancer controls, bridging insights from a high-risk group to the broader aging population. While epigenetic aging clocks will serve as our primary validation metric, we will cross-validate predictions against other established aging measures, including telomere length, frailty index, and cognitive function. Statistical validation will include correlation analyses to ensure biological and clinical relevance. To enhance interpretability, we will integrate deep learning for high-dimensional feature extraction, classical machine learning for explainability, and radiomics-based models for transparent biomarkers. This scalable, cost-effective framework can be immediately implemented in real-world settings, transforming routine clinical data into widely accessible aging biomarkers. By integrating AI with existing medical practice, our approach aligns with NAM’s mission to advance healthy longevity through bold, innovative solutions.