Comparison of hippocampal volume measurement by autopsy and post-mortem magnetic resonance imaging

Abstract

We present an autopsy-validated, non-invasive, magnetic resonance imaging (MRI) based segmentation algorithm, for determining hippocampal volume. A segmentation algorithm was developed to assess the volume of the hippocampus. Deceased individuals with severe mental illness were used to evaluate the use of MRI imaging to determine hippocampal volume as this group has previously been associated with altered hippocampal volume diagnosed on MRI. The accuracy of the MR- scanning protocol for volume measurement was tested on a water filled phantom control with a known volume of 500 ml, and a difference of 0.08% was found. Thus the scanning protocol was deemed to have produced acceptable results when comparing volume measures of a pair of segmented hippocampi obtained at the 1 T MR scanner and a 3 T MR scanner using the software program Mimics®. The segmentation algorithm was tested by a volume comparison obtained using anterior and posterior landmarks (in situ) and the exact volume of the dissected hippocampus (ex situ). The in situ and ex situ hippocampal volumes were highly correlated; R² was 96%, with a mean difference of 4-5%. Cases were also examined for intra- and inter-observer agreement. This study presents a validated segmentation algorithm that can be used to determine the hippocampal volume using post-mortem MR and anatomical landmarks.

Keywords: Autopsy; Hippocampus; Neurology; Post Morten magnetic resonance imaging; Segmentation.