Validation of hippocampal volumes measured using a manual method and two automated methods (FreeSurfer and IBASPM) in chronic major depressive disorder

Abstract

Introduction: To validate the usefulness of the packages available for automated hippocampal volumetry, we measured hippocampal volumes using one manual and two recently developed automated volumetric methods.

Methods: The study included T1-weighted magnetic resonance imaging (MRI) of 21 patients with chronic major depressive disorder (MDD) and 20 normal controls. Using coronal turbo field echo (TFE) MRI with a slice thickness of 1.3 mm, the hippocampal volumes were measured using three methods: manual volumetry, surface-based parcellation using FreeSurfer, and individual atlas-based volumetry using IBASPM. In addition, the intracranial cavity volume (ICV) was measured manually.

Results: The absolute left hippocampal volume of the patients with MDD measured using all three methods was significantly smaller than the left hippocampal volume of the normal controls (manual P = 0.029, FreeSurfer P = 0.035, IBASPM P = 0.018). After controlling for the ICV, except for the right hippocampal volume measured using FreeSurfer, both measured hippocampal volumes of the patients with MDD were significantly smaller than the measured hippocampal volumes of the normal controls (right manual P = 0.019, IBASPM P = 0.012; left manual P = 0.003, FreeSurfer P = 0.010, IBASPM P = 0.002),. In the intrarater reliability test, the intraclass correlation coefficients (ICCs) were all excellent (manual right 0.947, left 0.934; FreeSurfer right 1.000, left 1.000; IBASPM right 1.000, left 1.000). In the test of agreement between the volumetric methods, the ICCs were right 0.846 and left 0.848 (manual and FreeSurfer), and right 0.654 and left 0.717 (manual and IBASPM).

Conclusion: The automated hippocampal volumetric methods showed good agreement with manual hippocampal volumetry, but the volume measured using FreeSurfer was 35% larger and the agreement was questionable with IBASPM. Although the automated methods could detect hippocampal atrophy in the patients with MDD, the results indicate that manual hippocampal volumetry is still the gold standard, while the automated volumetric methods need to be improved.