Amygdalar nuclei and hippocampal subfields on MRI: Test‐retest reliability of automated segmentation in old and young healthy volunteers: Neuroimaging / Normal brain aging.

Background: The amygdala and the hippocampus are two limbic structures that play a critical role in cognition and behaviour but their small size hampers their manual segmentation in multicenter datasets. Here, we assessed the reliability of the automated segmentation of amygdalar nuclei and hippocampal subfields across sites and vendors. We applied a new high‐resolution atlas in two independent cohorts of older and younger healthy adults. Method: Sixty‐four older (PharmaCog study; age range: 50‐78 years) and 67 younger subjects (CoRR consortium; age range: 18‐43 years) underwent repeated 3D‐T1 MRI at two different time points between 1 and 90 days. Amygdala and hippocampus segmentation was performed using FreeSurfer v6.0. Reliability of functional parcellations was assessed using volume reproducibility error (ε) and spatial overlapping coefficient (DICE) between test and retest session. Nuclei/subfields were considered as single structures and grouped into relevant subregions (for the amygdala, basolateral and the centromedial complexes; for the hippocampus, head, body, and tail subdivisions). Result: Differences in MRI site/vendor had a significant impact (p<.05) for the ε of few subfields/nuclei, while extensively impacted DICE of most structures. The effect on DICE was limited by the functional parcellation only in the younger cohort. Reliability was strongly influenced by volume, as ε correlated negatively and DICE correlated positively with volume size of structures (|rs|>.43, p<.001). In line, nuclei larger than 200 mm3 and subfields larger than 300 mm3 (except for molecular layer) had the best test‐retest reproducibility (ε<5% and DICE>0.80). Conclusion: Our results support the use of volumetric measures derived automatically from larger amygdalar nuclei and hippocampal subfields as reliable measures in multisite longitudinal MRI studies. These measures could be useful for studies evaluating disease progression and treatment response. [ABSTRACT FROM AUTHOR]