Comparison of MR imaging against physical sectioning to estimate the volume of human cerebral compartments

• Published in: NeuroImage (Orlando, Fla.) Vol. 18; no. 2; pp. 505 - 516
• Main Authors: García-Fiñana, Marta, Cruz-Orive, Luis M., Mackay, Clare E., Pakkenberg, Bente, Roberts, Neil
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 01.02.2003; Elsevier Limited
• Subjects: Aged; Biomedical research; Brain; Cavalieri volume estimator; Cerebral Cortex - anatomy & histology; Dominance, Cerebral - physiology; Female; Formalin-fixed specimens; Gray matter; Humans; Image Processing, Computer-Assisted; Imaging, Three-Dimensional; Magnetic Resonance Imaging; Male; Methods; Middle Aged; MRI; Numerical Analysis, Computer-Assisted; Physical sectioning; Point counting; Reference Values; Sensitivity and Specificity; Software; Stereology; Studies; Systematic sampling; White matter; Cavalieri volume estimator; Physical sectioning; Systematic sampling; MRI; Formalin-fixed specimens; Point counting; Stereology; Gray matter; White matter
• ISSN: 1053-8119; 1095-9572
• DOI: 10.1016/S1053-8119(02)00021-6

The purpose of this study was to compare magnetic resonance imaging (MRI) against physical sectioning techniques to estimate the volume of human cerebral hemisphere compartments (cortex, subcortex, and their union, called “total”). The volume of these compartments was estimated postmortem for six human subjects from MRI virtual sections and from physical sections using the Cavalieri design with point counting. Cursory paired t tests revealed no significant differences between the two methods for any of the three compartments considered, although P = 0.06 for the subcortex. A sharper analysis incorporating recent error prediction formulae revealed a significant discrepancy between the two methods in the estimation of subcortex and total volume for three of the specimens. Yet, none of these analyses is adequate to detect possible biases. The incorporation of an explanatory variable, namely hemisphere weight, and the adoption of a specific gravity ρ = 1.04 g/cm 3 for the material, enabled us to carry out an allometric analysis for the total compartment which revealed a significant bias of the MRI data. The new error prediction formulae are illustrated by way of example, and their accuracy is checked by a resampling experiment on a data set of 274 MRI sections.