Inter-observer and inter-MRI sequence differences in the assessment of cortical strokes. Do they matter in research?

1. Department of Neuroimaging Sciences, Centre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, UK
2. College of Medicine and Veterinary Medicine, University of Edinburgh, Edinburgh, UK

Background: We analyse variability of inter-observer and inter-sequence differences in the assessment of cortical strokes(CS) from magnetic resonance images(MRI) and its possible influence in their association with clinical parameters.

Methods: Data were from 57 patients with CS. CS were delineated semi-automatically by two observers, blind to each-other, in MRI-FLAIR and MRI-T1-weighted(T1W), aided-by vs. blind-to DWI. We used Bland-Altman to analyse inter-observer and inter-sequence differences, and univariate linear regression for associations with clinical variables.

Results: The use of DWI identified on average 2.6ml(95%CI[-5.4 +10.5]ml) of additional volume in index CS and 1.1ml(95%CI[-6.4 +8.6]ml) in old CS compared to when only FLAIR was used. FLAIR identified on average 3.8ml(95%CI[-6.7 +14.4]ml)  of additional volume on recent CS and 8.8ml(95%CI[-5.1 +6.9]ml) on old CS compared to T1W. FLAIR vs.T1W discrepancies were mainly in the middle cerebral artery territory, comparable in both hemispheres. Inter-observer differences were mainly in the right posterior cerebral artery territory. Inter-sequence, but not inter-observer differences increased with the increase of the CS volume. None of the measurements was associated with the clinical parameters evaluated: age, basal ganglia perivascular spaces burden, blood pressure, pulse frequency, small vessel disease load, Fazekas or atrophy scores, and all associations yielded similar B and p values.

Conclusions: T1W is the least sensitive sequence for establishing the boundaries and extent of, mainly, old CS. In absence of DWI, inter-observer differences for recent CS are higher. On average, inter-observer and inter-sequence differences in the quantification of CS volume do not seem to determine their relationship with clinical parameters.

Funded by: Medical Research Scotland, Wellcome Trust, Row Fogo Charitable Trust