Ex vivo human brain volumetry: validation of magnetic resonance imaging measurements
BackgroundNeurodegenerative diseases are associated with brain atrophy. The volume of in vivo human brains is determined with various magnetic resonance imaging (MRI) measurement tools of which the validity has not been assessed against a gold standard. Here, we propose to validate the MRI brain volumes by scanning ex vivo-in situ specimens (i.e., anatomical heads), which allows the extraction of the brain after the scan to compare its volume with the gold standard water displacement method (WDM). MethodsWe acquired 3T MRI T2-weighted, T1-weighted, and MP2RAGE images of seven anatomical heads fixed with an alcohol-formaldehyde solution routinely used in anatomy laboratories and segmented the gray and white matter of the brain using two methods: 1) a manual intensity-based threshold segmentation using Display (MINC-ToolKit, McConnell BIC), and 2) an automatic Deep-Learning-based segmentation tool (SynthSeg). The brains were then extracted, and their volumes were measured with the WDM after the removal of their meninges and a midsagittal cut (to allow water penetration into the ventricles). Volumes from all methods were compared to the ground truth (WDM volumes) using a repeated-measures ANOVA. ResultsMean brain volumes, in cubic centimeters, were 1111.14{+/-}121.78 for WDM, 1020.29{+/-}70.01 for manual T2-weighted, 1056.29{+/-}90.54 for automatic T2-weighted, 1094.69{+/-}100.51 for automatic T1-weighted, 1066.56{+/-}96.52 for automatic MP2RAGE INV1, and 1156.18{+/-}121.87 for MP2RAGE INV2. All volumetry methods were significantly different (F=17.874; p<0.001) from the WDM volumes, except the automatic T1-weighted volumes. ConclusionWe demonstrate that SynthSeg accurately determines the brain volume in ex vivo-in situ T1-weighted MRI scans. Our results also suggest that given the contrast similarity between our ex vivo and in vivo sequences, the brain volumes of clinical studies are most probably sufficiently accurate, with some degree of underestimation depending on the sequence used.