Confidence map for multi-gradient-based X-ray phase-contrast imaging: a probabilistic error classification approach

Phase images recovered from multi-gradient-based X-ray phase-contrast imaging can exhibit several error modes, which makes assessing their reliability essential. To exhibit them, one approach is to use the phase gradients measurements to calculate the local phase-derivative closure, which may reveal inconsistencies caused by undersampling or noise. Our previous work [ Opt. Express 30 , 4302 ( 2022 ) 10.1364/OE.438876 ] introduced a graphical tool called Confidence Map to highlight these artifacts. Here, we introduce a new, to the best of our knowledge, version based on a Probabilistic Phase Error Classification . This probabilistic method converts closure deviations into per-pixel error probabilities and produces alert maps highlighting regions where the recovered phase may be unreliable. These maps give a compact, interpretable view of where phase extraction is trustworthy versus limited by sampling, and they provide actionable guidance for both acquisition design and downstream processing. We applied this approach to a carbon composite sample imaged with multilateral shearing interferometry, a single 2D-grating technique, showing its practicality and its ability to accurately localize artifacts.