This section describes the datasets used to optimise and validate BIANCA for the detection of white matter hyperintensities of presumed vascular origin (Wardlaw et al., 2013 (link)). The datasets are different in terms of populations, were acquired on different scanners and using different imaging protocols (see details below).
Exclusion criteria applied to both cohorts for the purposes of the present study were: presence of intracranial haemorrhage; intracranial space occupying lesion; WMH mimics (multiple sclerosis and irradiation induced gliosis); brain defect due to previous neurosurgery or developmental anomalies; large chronic, subacute or acute infarcts (i.e., > 2 cm on either T1-, T2-weighted or DWI sequences); significant movement artefacts.
For both datasets WMHs were graded on FLAIR images by a trained operator (L.L.) who provided visual ratings according to the following scales: 1) a modified version of the Fazekas scale (Fazekas et al., 1987 (link)), considering periventricular and deep white matter lesions altogether (range total score 0–6); 2) the ARWMC (Age-Related White Matter Changes, (Wahlund et al., 2001 (link))) scale, rating 5 different regions (frontal, parieto-occipital, temporal, basal ganglia, infratentorial) in both hemispheres according to a 0–3 score (range total score 0–30).
Free full text: Click here