National Institute on Aging 2R01AG070149-04

Multiorgan Photon Counting CT and Machine Learning to Elucidate Aging Mechanisms and Interventions

This NIH-funded project at Duke University uses photon-counting CT and machine learning to study how liver changes influence aging in multiple organs, focusing on genetic and lifestyle factors. It aims to develop a biological age biomarker and understand mechanisms by which diet and exercise affect aging, with potential for clinical translation.

Description

Aging is systemic yet organ-specific, and we lack in-vivo tools that track how genotype, diet, and exercise reshape multiorgan trajectories. The apolipoprotein E (APOE) gene is the most common genetic modifier of Alzheimer’s disease, cardiovascular disease, and osteoporosis. Our central hypothesis is that early shifts in hepatic fat fraction, iron content, and perfusion act upstream to modulate biological aging in the heart, brain, and bone. We propose a liver-centered model of systemic aging, leveraging photon-counting computed tomography (PCCT) and machine learning to quantify organ-specific age trajectories and cross-organ influence. We will study APOE2, APOE3, and APOE4 knock-in mice with or without a humanized NOS2 allele (hNOS2), exposed to control diet, high-fat diet (HFD), voluntary exercise, or HFD + exercise. Longitudinal in vivo and ex vivo PCCT scans will be acquired at 6, 12, and 18 months to quantify morphologic and perfusion changes in liver, heart, brain, and bone. Deep learning–based segmentation and radiomics will extract age-sensitive features, complemented by behavioral assessments of memory, motor function, and activity. We will train contrastive learning models to predict organ-specific biological age (ΔAge) from imaging and behavioral features and construct a Multiorgan Biological Age (MBA) clock using a graph neural network framework. The MBA model will quantify directional aging influences—particularly liver-to-organ pathways—and be validated against behavioral and histological outcomes. Molecular profiling (targeted RNA-seq and cytokines) will be integrated with imaging data to test whether hepatic normalization mediates systemic rejuvenation under exercise. Structural equation modeling and cross-lagged analyses will determine whether changes in liver composition and cytokine output causally explain exercise-induced reductions in ΔAge across organs. Expected outcomes include: (1) the first APOE-stratified, multimodal atlas linking hepatic phenotypes to systemic aging; (2) an imaging-assisted MBA biomarker deployable with clinical PCCT systems; and (3) mechanistic insights into how diet and exercise modulate aging through hepatic reprogramming. This project enables direct translation to human studies, positioning the liver as both a biomarker source and a therapeutic target to slow age-related disease.

View original record