Objectives: To analyze the texture features on cranial sonography in preterm neonates with white matter injury quantitatively and to correlate these features with magnetic resonance imaging (MRI).
Methods: The study included 33 preterm neonates treated in our neonatal intensive care unit who underwent serial cranial sonography and brain MRI near term. Patients were subdivided into 3 groups according to the presence and severity of white matter injury as revealed by MRI: normal (group 1; n = 20), mild (group 2; n = 5), and severe (group 3; n = 8). The periventricular echogenicity on sonography was evaluated quantitatively with second-order gray-level statistics (gray-level co-occurrence matrix [GLCM] method). Four GLCM texture features representing homogeneity were extracted in 12 directions: (1) angular second moment (ASM), (2) inverse differential moment (IDM), (3) contrast, and (4) entropy.
Results: Thirty of 48 features showed a statistically significant difference between groups 1 and 3 (ASM in 9 directions, IDM in 6 directions, contrast in 3 directions, and entropy in all 12 directions). There were no significant differences observed between groups 1 and 2 or groups 2 and 3. The mean contrast and entropy values were generally lower in group 1 than group 3, whereas the mean ASM and IDM values were higher in group 1.
Conclusions: Severe white matter injury could be identified by using GLCM texture analysis, whereas mild white matter injury observed on MRI could not be evaluated by GLCM analysis. Quantitative texture analysis using the GLCM may serve as a complementary tool for quantitative assessment of periventricular echogenicity.
Keywords: gray-level co-occurrence matrix; pediatric ultrasound; periventricular leukomalacia; quantitative analysis; sonography; white matter injury.
© 2015 by the American Institute of Ultrasound in Medicine.