Multiscale partial volume estimation for segmentation of white matter lesions using flair MRI

For robust segmentation of white matter lesions (WML), a partial volume fraction (PVF) estimation approach was previously developed for FLAIR MRI that does not depend on predetermined intensity distribution models or multispectral scans. Instead the PV fraction was estimated directly from each FLAIR MRI using an adaptively defined global edge map that exploits a novel relationship between edge content and PVA. Although promising, predefined noise filter parameters were needed, and the edge metric is computed on a single scale which limits wide-scale implementations. To handle these challenges, this work defines a novel multiscale PVF estimation approach that is based on scale space derivatives. The result is a scale-invariant representation of edge content which is used to estimate a multiscale (scale-invariant) PV fraction. Validation results show the method is performing better than the previous version.