Background: Over 1 million infants born annually with fetal exposure to HIV and maternal antiretroviral treatment (ART) who remain HIV-uninfected (HEU) are at higher risk of neurodevelopmental delays compared to infants HIV-unexposed (HU).
Objective: We explored the use of brain ultrasound radiomics, specifically texture analysis, as an early imaging neurodevelopmental biomarker, comparing findings by newborn in utero HIV exposure status.
Methods: Brain ultrasound was performed on full-term newborns (≥ 37 weeks gestation) enrolled in a prospective observational study in Botswana. Radiomic ultrasound features, including first-order statistics, run-length, and co-occurrence matrix parameters, were extracted from the basal ganglia and periventricular white matter. Statistical comparisons were conducted based on fetal exposure to maternal HIV. The diagnostic performance of individual features was assessed, and logistic regression was used to combine the features for overall performance evaluation.
Results: Thirty-three infants (HEU: 20, HU: 13) were included in the analysis. The basal ganglia of HEU infants exhibited significantly lower heterogeneity (176.6 ± 10.76 vs. 205.97 ± 13.26, p = 0.04) and entropy (0.37 ± 0.01 vs. 0.41 ± 0.01, p = 0.03), and marginally lower gray level non-uniformity (310.04 ± 15.32 vs. 352.37 ± 24.20, p = 0.06) compared to HU infants, suggesting reduced parenchymal complexity. These combined radiomic features yielded an AUC of 0.72 with a specificity of 0.86. Similar trends were observed in the white matter, where HEU infants demonstrated marginally lower heterogeneity (191.66 ± 14.32 vs. 231.76 ± 17.34, p = 0.06). Gray level non-uniformity and run length non-uniformity were significantly lower in the HEU group (1996.87 ± 157.06 vs. 2487.43 ± 223.67, p = 0.04 and 284.66 ± 20.37 vs. 406.61 ± 47.77, p = 0.01, respectively). The combined white matter model demonstrated an AUC of 0.76 and a sensitivity of 0.86, indicating greater discriminatory power compared to the basal ganglia.
Conclusion: Ultrasound radiomics reveals distinct differences in brain texture between HEU and HU newborns, with significant findings in both basal ganglia and white matter features. These results highlight the potential of radiomics in identifying subtle neuroanatomical variations. Further research is needed to explore the neurodevelopmental implications of these findings.
Objectives: The objectives of this study were to determine whether infants born to mothers with gestational or pregestational diabetes (IDM) and infants born to mothers with uncomplicated pregnancies have differences in percent body fat at birth, and to examine the relationship between glycemic control in pregnancy and infant body fat.
Study design: Percent body fat was measured in 132 IDM in the well-baby nursery using air displacement plethysmography and compared to a previously published cohort of 249 healthy control infants using t-tests. Linear regression models were used to examine the relationship between maternal glycemic control and infant percent body fat at birth.
Results: The difference in body fat between IDM and infants born to mothers with uncomplicated pregnancies was not clinically significant (12.9 ± 5 % vs 13.9 ± 4 %, p < 0.05). Among IDM, increasing maternal fasting blood glucose values and increasing percentages of elevated post-prandial glucose values were associated with increasing infant percent body fat. (β=0.19 and 0.08 respectively, P < 0.05).
Conclusion: Our cohort of healthy IDM that remained admitted to the term nursery did not have a clinically significant difference in percent body fat compared to control infants born to mothers with uncomplicated pregnancies. Among IDM, elevated maternal glucose levels correlated with increased infant percent body fat, suggesting glycemic control in pregnancy can affect body composition within this population.

