Effect of surgery on periventricular white matter in normal pressure hydrocephalus patients: comparison of two methods of DTI analysis.

ABSTRACT

BACKGROUND: Diffusion tensor imaging (DTI) is a useful tool for assessing changes that occur in microstructures. We have developed a novel method for region of interest (ROI) delineation in the assessment of DTI parameters in patients with normal pressure hydrocephalus (NPH). PURPOSE: To compare the standard method and our novel method in an evaluation of the impact of surgery on periventricular white matter in patients with NPH. MATERIAL AND METHODS: Ten patients with NPH underwent 3T magnetic resonance imaging (MRI; including 12-direction DTI sequences) before and after surgery. We recorded diffusion parameters (λi, the fractional anisotropy [FA], the apparent diffusion coefficient, and Dr) in the internal capsule (IC) and the body of the corpus callosum (BCC). Using the standard delineation technique, regions of interest (ROIs) were positioned according to anatomical and functional considerations and then filled with several sub-ROIs. The ROIs delineated with our novel technique (extracted as the six sub-ROIs with the lowest standard deviation for the FA) were arranged in two rows (medial and lateral), from the ventricle to the brain surface. RESULTS: The within-ROI homogeneity was higher with the novel method than with the conventional method (P<10(-4)). When the conventional delineation method was applied to the IC data, only λ2 was found to be significantly greater after surgery; in contrast, application of our novel method evidenced a significant decrease in FA and λ1 and a significant increase in λ2 (P<0.05). Both before and after surgery, the FA in the medial row of ROIs was greater than the FA in the lateral row (P<0.01). In the BCC, only λ2 and Dr varied significantly (when evaluated with the novel method). CONCLUSION: Our results show that use of a novel method of DTI data analysis may be more sensitive to local changes induced by surgical procedures. Furthermore, this novel method was able to detect the transmantle pressure gradient related to the regional stress distribution.