Quantifying model uncertainty for semantic segmentation of Fluorine-19 MRI using stochastic gradient MCMC

Masoumeh Javanbakhat, Ludger Starke, Sonia Waiczies, Christoph Lippert

Abstract: Highlights•Fluorine-19 (19F) MRI is a promising tool for studying diseases and treatments.•Analysis of 19F MRI data necessitates detecting the 19F MRI signals from noise.•Current methods use background subtraction, but suffer from low sensitivity.•BDL enhances 19F MRI sensitivity significantly, generating reliable uncertainty maps.•Uncertainty maps aid 19F experts to analyze predictions, thus enhancing reliability.