A new brain-imaging model claims it can estimate a person's biological brain age more accurately than before, across nearly the entire human lifespan. The catch: how much more accurate depends entirely on which older method you compare it to.
Researchers built a two-stage framework that combines two types of MRI data, brain structure and white matter organization, processed separately and then merged. The first stage sorts a scan into one of six life stages, from fetal to elderly. The second stage then predicts an exact age using specialized models weighted by that stage classification. Tested on nine datasets spanning fetus to old age, the method cut prediction error by 13 percent against one existing baseline and by 78 percent against another, with combining both MRI types adding a further 12-13 percent improvement on its own.
That 78 percent figure is the headline number, but it is a comparison against one weaker baseline, not a blanket verdict over prior methods generally. The more meaningful result may be a follow-on analysis using ADNI Alzheimer's data, where the gap between predicted and actual brain age tracked with disease-related aging patterns, hinting at a usable biomarker.
Brain-age gap has been studied as an aging proxy for years; what is new here is stitching structure and connectivity data together across the full lifespan instead of one narrow age band. Still, this is a preprint, and papers tend to lead with their best number, not their most representative one.