AUTHOR=Zhang Chunxiang , Zhu Zitao , Wang Kaiyu , Wang Linlin , Lu Jiaqi , Lu Lin , Xing Qingna , Wang Xueyuan , Zhang Xiaoan , Zhao Xin 

TITLE=Predicting neurodevelopmental outcomes in extremely preterm neonates with low-grade germinal matrix-intraventricular hemorrhage using synthetic MRI

JOURNAL=Frontiers in Neuroscience

VOLUME=18

YEAR=2024

URL=https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2024.1386340

DOI=10.3389/fnins.2024.1386340

ISSN=1662-453X

ABSTRACT=<sec id="sec1"><title>Objectives</title><p>This study aims to assess the predictive capability of synthetic MRI in assessing neurodevelopmental outcomes for extremely preterm neonates with low-grade Germinal Matrix-Intraventricular Hemorrhage (GMH-IVH). The study also investigates the potential enhancement of predictive performance by combining relaxation times from different brain regions.</p></sec><sec id="sec2"><title>Materials and methods</title><p>In this prospective study, 80 extremely preterm neonates with GMH-IVH underwent synthetic MRI around 38 weeks, between January 2020 and June 2022. Neurodevelopmental assessments at 18 months of corrected age categorized the infants into two groups: those without disability (<italic>n</italic> = 40) and those with disability (<italic>n</italic> = 40), with cognitive and motor outcome scores recorded. T<sub>1</sub>, T<sub>2</sub> relaxation times, and Proton Density (PD) values were measured in different brain regions. Logistic regression analysis was utilized to correlate MRI values with neurodevelopmental outcome scores. Synthetic MRI metrics linked to disability were identified, and combined models with independent predictors were established. The predictability of synthetic MRI metrics in different brain regions and their combinations were evaluated and compared with internal validation using bootstrap resampling.</p></sec><sec id="sec3"><title>Results</title><p>Elevated T<sub>1</sub> and T<sub>2</sub> relaxation times in the frontal white matter (FWM) and caudate were significantly associated with disability (<italic>p</italic> &lt; 0.05). The T<sub>1</sub>-FWM, T<sub>1</sub>-Caudate, T<sub>2</sub>-FWM, and T<sub>2</sub>-Caudate models exhibited overall predictive performance with AUC values of 0.751, 0.695, 0.856, and 0.872, respectively. Combining these models into T<sub>1</sub>-FWM + T<sub>1</sub>-Caudate + T<sub>2</sub>-FWM + T<sub>2</sub>-Caudate resulted in an improved AUC of 0.955, surpassing individual models (<italic>p</italic> &lt; 0.05). Bootstrap resampling confirmed the validity of the models.</p></sec><sec id="sec4"><title>Conclusion</title><p>Synthetic MRI proves effective in early predicting adverse outcomes in extremely preterm infants with GMH-IVH. The combination of T<sub>1</sub>-FWM + T<sub>1</sub>-Caudate + T<sub>2</sub>-FWM + T<sub>2</sub>-Caudate further enhances predictive accuracy, offering valuable insights for early intervention strategies.</p></sec>