Cerebral microbleed detection on 7T MRI: Reliability and effects of image processing

Abstract

7T MRI allows detection of smaller cerebral microbleeds than MRI at lower field strength. However, susceptibility effects of other structures do also complicate microbleed detection at higher field strength. Image processing by means of minimum intensity projection (minIP) is required for visual rating because of the large number of slices in high resolution scans. The present study determined optimal minIP conditions to detect susceptibility artefacts of microbleeds on 7T MRI, based on sensitivity of detection and rating time. Furthermore, intra- and inter-rater reliability of this visual microbleed detection on 7T MRI were assessed. 3D T2*-weighted images were acquired on a 7T MR system and scored by using the Microbleed Anatomical Rating Scale. First, eight different conditions of minIP processing, involving variations of slice thickness and gap, were applied to scans from three subjects with multiple microbleeds. Two of these conditions (slice thickness/gap in mm: 2/0, 4/-2) were applied to 7T scans of ten subjects to assess intra- and inter-rater reliability, with four raters, one of whom scored all scans twice. We verified that inter-rater reliability of these raters for visual scoring of regular, 1.5T, 3D T2*-weighted scans of the same subjects was within the range reported in previous studies. Of the eight conditions, 2/0 and 4/-2 minIP showed a good and comparable tradeoff between sensitivity for detection and rating time. Susceptibility artefacts of microbleeds from 0.7 mm in diameter or more could be detected on 7T scans (See figure for example). More microbleeds were detected on 7T (median (range): 2 (0-11)) than 1.5T scans (0 (0-3)). Intra-rater reliability for detection of number of microbleeds on 7T was good (Intraclass Correlation Coefficient: 0.73), but inter-rater reliability was only moderate (0.46).