Objective:

To explore the predictive value of whole-brain and region-specific Z-scored Apparent Diffusion Coefficient (ZADC) abnormalities in gray matter (GM) and white matter (WM) as early neuroimaging markers for neurodevelopmental outcomes at two years in neonates with Hypoxic-Ischemic Encephalopathy (HIE).

Background:

Brain injury in HIE is characterized by a low ADC, which is associated with poor neurodevelopmental outcomes. Previous studies often used fixed regional thresholds, disregarding anatomical heterogeneity and the continuum of diffusion metrics. This study utilizes ZADC maps, which normalize voxel-wise ADC to region- and tissue-specific norms from a healthy neonatal cohort. This Z-score transformation provides a uniform scale for quantifying abnormal diffusion and injury severity across different brain tissues.

Design/Methods:

We analyzed 21 neonates with HIE (mean GA 38.23 weeks, 18 males, 62% with seizures, 62% treated with therapeutic hypothermia). Patient ADC maps were used to create ZADC maps by comparison to a normative ADC atlas. Z-scores were calculated for whole-brain and specific regions of interest in GM and WM, defined by the M-CRIB2.0 atlas. Multivariate regression was used to predict Bayley-4 domain scores at two years using regional ZADC values.

Results:

Significant correlations were found between regional ZADC values and neurodevelopmental outcomes. In GM, right lentiform nucleus ZADC significantly predicted Language Composite (p=0.004), Expressive Communication (p=0.001), and Gross Motor (p=0.001) scores. ZADC in the right occipital lobe, right hippocampus, and cerebellum individually predicted Fine Motor scores (p<0.05). In WM, ZADC in the left occipital lobe was a significant predictor of Cognitive Composite (p=0.018) and Language Composite (p=0.024) scores.

Conclusions:

ZADC deviations in key GM and WM regions may provide early biomarkers of motor, language, and cognitive impairments at two years in HIE neonates. This continuous, voxel-based analysis offers a more nuanced approach for prognostication. A small sample size was the primary limitation, and future multi-center collaborations are warranted.