Novel Mammillary Body Manual Segmentation: Application for Quantitative MRI Analysis of Critically Ill Infants.

Abstract

Background and purpose: Previous qualitative studies have shown that mammillary body (MB) assessment can serve as an early marker of poor long-term neurodevelopmental outcomes. This study aims to establish a reliable quantitative method for analyzing the surface area, volume, and signal intensity of MB in infancy.

Methods: A novel methodology was retrospectively tested in a cohort of critically ill preterm and term-born patients following long-gap esophageal atresia (EA) repair, and healthy term-born controls (n = 13/group) using non-sedated brain MRI on a 3T Siemens scanner. Manual bilateral MB segmentation of T2-weighted data and quantification of MB surface area, volume, and tissue mean signal intensity were performed using ITK-SNAP. Endpoint measures were assessed for normality, and their relationship with group status was evaluated using a general linear model with age at scan as a covariate.

Results: High intra- and inter-tracer reliability was observed between a novice and neuroanatomical expert for MB segmentation. Despite straightforward manual masking and novel quantification of infant MB, no significant differences were found among the three groups (preterm and term-born patients, and term-born controls) for any of the MB endpoints analyzed: surface area, volume, and signal intensity. The data analysis revealed a trend of lower values in patient groups for signal intensity only.

Conclusions: This novel study describes efficient and accurate MB masking and quantification, supporting MB as a potential early marker. However, the negative results presented in infants born with long-gap EA should not be generalized until future prospective studies with larger sample sizes are conducted and linked to neurodevelopmental outcomes.