89 White matter hyperintensity shape is associated with long-term dementia risk 5 calculated using the box counting technique for both periventricular/confluent and deep WMHs. Lower convexity and solidity and a higher concavity index and fractal dimension indicate a more irregularly shaped WMH.18 In addition to fractal dimension, eccentricity was calculated for deep WMHs. Eccentricity was calculated by dividing the minor axis of a lesion by the major axis of the lesion and describes the deviation from a sphere. A lower eccentricity corresponds to a more elongated lesion, while a higher eccentricity corresponds to a rounder lesion.18 A more complex shape of a deep WMH is described by a higher fractal dimension.18 Table S.5.8.1 contains an overview of the shape markers. Mean shape values across all WMHs per participant were determined for periventricular/confluent WMHs, and for deep WMHs, shape values were calculated per WMH and then averages were calculated per participant.18 Volumes of periventricular/confluent, deep, and total WMHs were calculated within the shape pipeline. Intracranial volume was calculated by adding the volumes of gray matter, white matter, cerebrospinal fluid, and WMH.23 Gray matter, white matter, cerebrospinal fluid, and WMH were segmented automatically with a modified algorithm based on the Montreal Neurological Institute pipeline.24 The researchers were blinded for participant characteristics, including dementia outcome at follow-up. Visual quality checks were performed on original and processed MRI images. Exclusions were performed in a consensus meeting between involved researchers, including a neuroradiologist. 5.3.4 Analytical sample The inclusion and exclusion of participants from the AGES-Reykjavik study for the current study are illustrated in Figure S1. Participants were excluded for further analysis if their MRI images contained WMH oversegmentation (n = 124), incorrect ventricle segmentation (n = 4), artifacts (n = 30), brain infarcts > 15 mm (n = 460), tumors (n = 12), technical errors (n = 7), incorrect segmentations (n = 6), or traumatic brain injury (n = 2). A small group of participants (n = 62) had partial oversegmentation of WMHs besides (multiple) correct WMH segmentations. These participants were not excluded from further analysis. In addition, participants who were diagnosed with dementia at baseline (n = 137) were excluded for further analysis, because the present study focused on studying markers for developing dementia in the future. Participants with missing cognitive data at baseline (n = 81) or missing data at followup (n = 674) were also excluded for further analysis in this study. A final sample of n = 3077 was included in the final analysis (mean age: 75.6 ± 5.2 years).
RkJQdWJsaXNoZXIy MTk4NDMw