6533b7d0fe1ef96bd125ad51

RESEARCH PRODUCT

Segmentation of MR brain images with bias artifact

Edoardo ArdizzoneOrazio GambinoFrancesco AlagnaRoberto Pirrone

subject

Settore ING-INF/05 - Sistemi Di Elaborazione Delle InformazioniArtifact (error)Computer sciencebusiness.industryLow-pass filterScale-space segmentationPattern recognitionFilter (signal processing)Image segmentationMedical imagingBrain segmentationRF-Inhomogeneity bias segmentation Halo MRlComputer visionSegmentationArtificial intelligencebusiness

description

Brain MR Images corrupted by RF- Inhomogeneity (bias artifact) exhibit brightness variations across the image. As a consequence, a standard Fuzzy C-Means (fern) segmentation algorithm may fail. In this work we show a new general-purpose bias removing algorithm, which can be used as a pre-processing step for a fern segmentation. We also compare our experimental results with the ones achieved by using E2 D - H U M filter, showing an improvement in brain segmentation and bias removal.

yearjournalcountryeditionlanguage
2009-11-01
10.1109/itab.2009.5394449http://hdl.handle.net/10447/76935