MiniMORPH: A Morphometry Pipeline for Low-Field MRI in Infants

Abstract Background Ultra-low-field (ULF) MRI facilitates neuroimaging access, yet its application in early infancy is constrained by low resolution and contrast, and the limited suitability of existing segmentation tools. Objective To introduce and validate miniMORPH, an open-source pipeline for automated brain volumetry from 0.064T T2-weighted MRI acquired across infancy and toddlerhood. Methods ULF scans were acquired from infants aged 2 to 27 months across two cohorts in South Africa and Uganda. Age-specific templates and priors were used to segment major brain tissues and substructures. Validation used two high-field (HF) references: (i) expert manual HF segmentations for key ROIs across ages, and (ii) automated HF segmentations from SuperSynth on paired HF-ULF scans. We quantified (a) between-subject ordering across modalities using Pearson’s correlation (r) and (b) systematic scaling differences using percentage error (PE) and time-corrected percentage error (CPE), stratifying performance by cohort and age. Face validity was also tested via mixed-effects models of age, sex, and birthweight. Results miniMORPH successfully segmented major brain regions across infancy. In paired HF-ULF comparisons, between-subject ordering was generally preserved across many ROIs, with stronger correspondence in the South African cohort than in the Ugandan cohort at 12 months. Systematic scaling offsets were most evident in CSF-rich or boundary-sensitive compartments, with consistently negative CPE for ventricles and cerebellum. Performance varied with age, showing the greatest variability at 3 months. miniMORPH successfully captured regional age-related growth trajectories. Sex-dependent volumetric differences were widespread but attenuated after intracranial volume correction. Low birthweight infants exhibited reduced regional volumes and altered growth trajectories. Conclusion miniMORPH enables volumetric analysis of ULF infant MRI and preserves between-subject variation suitable for developmental and group analyses. ROI- and cohort-specific offsets, particularly in CSF-rich regions, may require calibration when absolute volumes are needed. The pipeline is openly available at github.com .