Using local binary pattern to classify dementia in MRI

White matter lesions (WML) are hyperintense signals in T2-weighted MRI of the brain. Volume and regional distribution of WML have been extensively studied in dementia, but not much attention has been given to texture analysis in these regions. We wanted to explore if it is possible to distinguish patients with dementia from healthy elderly in a classification framework testing different texture features in the WML regions, paying special attention to a feature called Local Binary Pattern (LBP). The results presented here, indicates that the LBP features used in our experiment are powerful features in a maximum likelihood classifier, when classifying demented from normal controls.