Differences in hippocampal volume developmental trajectory by processing pipeline

• In Event: Functional and structural findings from the Baby Connectome Project

Abstract

Introduction: It is well known that the hippocampus plays a critical role in learning and memory across the lifespan. Thus, to fully understand the development of learning and memory from infancy to childhood and beyond, it may be necessary to understand the developmental trajectory of the hippocampus. Previous studies on the developmental trajectory of hippocampal volumes have been small or cross-sectional, and larger, densely sampled studies of this trajectory are still needed.

Methods: Hippocampal volumes were generated from the processed data of two large multicenter studies, the Baby Connectome Project (BCP, 561 scans across 213 infants), and the Infant Brain Imaging Study (IBIS, 1373 scans across 623 infants). A subset of the BCP data was also processed using modified IBIS pipelines (BCP - IBIS processed, 341 scans across 198 infants). Hippocampal volumes by age were plotted for each, and Spearman’s rank correlation was calculated for the BCP and BCP - IBIS processed data.

Results: Dramatic differences in hippocampal volume trajectories were found between the BCP and IBIS studies (Figure 1). By 6 months old, the BCP volumes were on average 2353mm3 (55.8%) larger than the IBIS volumes. The trajectories from the IBIS volumes and the BCP - IBIS processed volumes were found to be comparable, suggesting that the differences seen between studies stemmed from processing differences. For each data point though, there was rank order consistency between the BCP data and the BCP - IBIS processed data, with a Spearman’s rank correlation of 0.93 (p<0.001), suggesting that these processing differences resulted mostly in a volume scaling shift. Upon further inspection, the biggest shift in the magnitude of difference between volumes of the two processing pipelines occurred near 5 months of age (Figure 2), corresponding to and possibly related to the contrast spin inversion that occurs in MRI images at this age.

Discussion: These findings show the impact that different processing pipelines can have on data and demonstrate the need for field-wide consensus on best practices in developmental image processing. As part of this, work is currently ongoing to create age-specific infant brain segmentations to share as a resource with the wider community.

Authors

• Sally Stoyell, University of Minnesota - Twin Cities
  Presenting Author

• Eric Feczko, University of Minnesota

• Trevor K.M. Day, University of Minnesota - Twin Cities

• Audrey Houghton, University of Minnesota

• Lucille A Moore, University of Minnesota - Twin Cities

• Joseph Piven, University of North Carolina at Chapel Hill

• Sun Hyung Kim, University of North Carolina

• Mark D. Shen, University of North Carolina at Chapel Hill

• Martin Styner, University of North Carolina at Chapel Hill

• Damien A. Fair, University of Minnesota - Twin Cities

• Jed T. Elison, University of Minnesota - Twin Cities