Researchers from the University of Hawaii hypothesized that total-body DEXA scans provide relevant body composition information beyond what's reflected in traditional risk factors. Furthermore, they theorized that tracking these body composition changes over time could improve mortality risk models.
They tested these hypotheses using deep-learning algorithms trained on data from 3,075 participants in the Health, Aging and Body Composition (Health ABC) study. Three models were trained: one using only analysis of changes in body composition from baseline and sequential total-body DEXA exams; one using only traditional mortality risk factors such as blood markers, general fitness indicators, and disability; and an algorithm that combined analysis of both DEXA information and traditional risk factors.
After testing each of the algorithms on a separate test set of cases, the researchers concluded that both of their hypotheses were confirmed.
"This approach provides a powerful way to study overall change in body composition characteristics coupled to clinical risk factors, paving a way for better intervention strategies to prolong lifespan," the authors wrote.
Attend this talk by presenter Yannik Glaser, PhD, to find out just how well the combined model performed for assessing mortality risk.