Purpose:
To compare manual region of interest (ROI) labeling and tract-based spatial statistics (TBSS) by their ability to detect group-wise differences in fractional anisotropy (FA) in the neonatal brain.
Materials and Methods:
Diffusion weighted data were obtained for nine infants with hypoxic-ischemic encephalopathy (HIE) (6 males, 3 females; gestational age range, 36-40 weeks; mean gestational age, 37.8 weeks) and eleven healthy-born infants (10 males, 1 female; gestational age range, 36-40 weeks; mean gestational age, 38.4 weeks) on a 3T scanner. For manual ROI labeling, eight ROIs were drawn freehand for each subject. For TBSS, all FA data underwent an optimized, automated protocol for neonates. Each method was evaluated for detection of decreased FA in HIE infants, sensitivity, specificity, and variability.
Results:
FA values from manual ROI and TBSS were strongly correlated. Both methods found decreased FA in most ROIs for HIE infants. There was no significant interaction between method and group, indicating a similar ability to detect FA differences. Sensitivity (manual: 0.71, TBSS: 0.69), specificity (manual and TBSS: 0.72), and standard error (manual: 0.009, TBSS: 0.007) were comparable.
Conclusions:
Manual ROI labeling and TBSS are comparable methods of diffusion analysis to detect group differences in FA in the neonatal brain