A method for rapid volumetric analysis of structural magnetic resonance images of the brain

Abstract

Objectives: To describe the methodology for volumetric analyses of brain volumes in vivo by structural magnetic resonance imaging. To assess the reliability and validity of the volumetric technique. Methods: Patients were recruited for magnetic resonance imaging as part of a wider project studying biological determinants of psychosis. Volumetric analysis of brain scans was performed blind using a rapid automated in-house software package to remove non-brain elements (scalp-editing), and to calculate volumes of the whole brain, lateral ventricles, cerebrospinal fluid, and cortical gray and white matter compartments. Results: Quantitative analyses of each magnetic resonance imaging scan took approximately 45 minutes for each patient (between-group results will be reported later). Test-retest reliability for 20% of scans selected at random was generally high. Spearman's rank correlation coefficient (rho) for whole brain was 0.99, grey matter was 0.79, white matter was 0.86, cerebrospinal fluid and sulcal compartment was 0.83 and lateral ventricle was 0.86. Inter-rater reliability by two independent assessors for the same sample of scans was high at 0.99 for whole brain and moderate for the other measures (grey matter and white matter was 0.62, cerebrospinal fluid and sulcal compartment was 0.60, lateral ventricle was 0.67). Face validity of the three-dimensional brain was good. An independent phantom analysis (measuring the volume of an object of known volume and intensity compartments) suggested that the estimated volumetric techniques were accurate to 97% and 98%, respectively. Conclusion: This method is reliable, valid, and fast for the purpose of quantification of cerebral morphology on magnetic resonance imaging scans. We propose that it can be employed for the assessment and monitoring of neuropsychiatric disorders in which cerebral volumetric change occurs.