Bayesian framework reveals uncertainty in gray matter diffusion MRI microstructural estimates

A Bayesian deep learning framework (µGUIDE) applied to gray matter diffusion MRI models shows that exchange time and soma radius estimates are often unreliable, while extracellular diffusivity and neurite signal fraction are robust. Uncertainty-aware methods help flag poor estim…

• In simulated and in vivo data, extra-cellular diffusivity and neurite signal fraction from NEXI and SANDIX models were robustly estimated, but exchange time and soma radius showed high uncertainty and bias under realistic noise and shorter acquisition protocols.
• The Bayesian approach outperformed conventional least squares fitting by flagging unreliable estimates, highlighting the value of uncertainty quantification for clinical translation.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.