Brain parenchymal fraction in an age-stratified healthy population : determined by MRI using manual segmentation and three automated segmentation methods

BACKGROUND AND PURPOSE: Brain atrophy is a prominent feature in many neurodegenerative diseases, such as multiple sclerosis, but age-related decrease of brain volume occurs regardless of pathological neurodegeneration. Changes in brain volume can be described by use of the brain parenchymal fraction (BPF), most often defined as the ratio of total brain parenchyma to total intracranial space. The BPF is of interest both in research and in clinical practice. To be able to properly interpret this variable, the normal range of BPF must be known. The objective of this study is to present normal values for BPF, stratified by age, and compare manual BPF measurement to three automated methods. MATERIALS AND METHODS: The BPFs of 106 healthy individuals aged 21 to 85 years were determined by the automated segmentation methods SyMap, VBM8 and SPM12. In a subgroup of 54 randomly selected individuals, the BPF was also determined by manual segmentation. RESULTS: The median (IQR) BPFs of the whole study population were 0.857 (0.064), 0.819 (0.028) and 0.784 (0.073) determined by SyMap, VBM8 and SPM12, respectively. The BPF decreased with increasing age. The correlation coefficients between manual segmentation and SyMap, VBM8 and SPM12 were 0.93 (P<0.001), 0.77 (P<0.001) and 0.56 (P<0.001), respectively. CONCLUSIONS: There was a clear relationship between increasing age and decreasing BPF. Knowledge of the range of normal BPF in relation to age group will help in the interpretation of BPF data. The automated segmentation methods displayed varying degrees of similarity to the manual reference, with SyMap being the most similar.