Automatic segmentation of age-related white matter changes on flair images: Method and multicentre validation

White matter hyperintensities (WMH) are commonly seen on T2-weighted images in elderly people. They are considered as a potential marker of vascular pathology and have been associated with motor and cognitive deficits. In this paper, non linear diffusion was applied to FLAIR images and combined with precise anatomical knowledge extracted from T1-weighted images to automatically segment WMH. Evaluation was performed on 24 patients with mild cognitive impairment (MCI) from 5 different centres. Results showed excellent volume agreement with manual delineation (Pearson coefficient: r=0.98, p<;0.001) and substantial spatial correspondence (Similarity index: 66%±17%). Our method appeared robust to acquisition differences across the centres.