595 Correlation Between Ultrasound Texture Analysis of the Cerebellum and Standard Diagnosis of White Matter Damage in Preterm Neonates

Abstract

Background and aims: Ultrasound diagnosis of white matter damage (WMD) is still subject to variability, and methods based on quantitative ultrasound analysis are under investigation to improve diagnostic accuracy. The correlation between WMD and cerebellar changes in term and one-year old infants has previously been described. We tested the hypothesis that a semiautomated method for ultrasound texture analysis (AQUA) could detect patterns to predict the standard clinical diagnosis of WMD in preterm neonates. Methods: A cohort of 40 very preterm infants (< 32 weeks) underwent 2 cranial US scans within the first and after 2 weeks of life. WMD was diagnosed using standard criteria. Texture analysis was run in the cerebellum, manually delineated in a coronal plane through the anterior fontanelle. A classification algorithm based on texture coefficients was used to correlate to clinical diagnosis and its ability to predict in the first scan a later diagnosis of WMD was then evaluated. Results: Median gestational age was 29+4 weeks (26+0-31+6). WMD was diagnosed in 10/40 infants. Cerebellar texture analysis in the first scan could identify cases later diagnosed with PVL with a sensitivity of 100% and an accuracy of 95%, after adjustment for gestational age and birth weight. Conclusions: The cerebellum might contain useful information that improves the prediction of standard white matter damage diagnosis. These preliminary data support further investigation of the role of this brain structure in the neonatal diagnosis of WMD by means of semi-automated quantitative ultrasound analysis.