Distribution Characteristics, Reproducibility, and Precision of Region of Interest–Based Hippocampal Diffusion Tensor Imaging Measures

Abstract

BACKGROUND AND PURPOSE: For adequate interpretation of diffusion tensor imaging (DTI) parameters empirical distribution characteristics, precision, and reproducibility should be known. The present study investigated distribution and reliability parameters of hippocampal fractional anisotropy (FA) and mean diffusivity (MD).

METHODS: FA and MD values were averaged in hippocampal regions of interest in 20 subjects (10 women and 10 men; age range, 25–69 years). Regions of interest were manually placed bilaterally by one investigator at 2 occasions, and by a second independent investigator. Sample distributions of FA and MD values were compared with normal distributions. Intraclass coefficients (ICCs), standard errors of measurement (SEMs), and coefficients of variation (CVs) with confidence intervals (CI95s) were computed.

RESULTS: The results did not show any deviation of averaged FA (0.237 ± 0.017) and MD (775 ± 28 μm²/s) values from normal distribution. Intraobserver reliability (ICC ≥ 0.90) and precision (CV ≤ 3.5%) were high for all measures. Interobserver reliability reached values of ICC ≥ 0.84 and CV ≤ 4.1%. FA yielded lower precision (CV 2.2–4.1%) than MD (CV 1.3–2.5%), CI95s were around ±0.015–0.020 and ±25–30 μm²/s for FA and MD, respectively. FA differences of 0.020–0.030 and MD differences of 40–50 μm²/s can be assumed to reflect reliably distinct values in hippocampal regions.

CONCLUSION: The results are in line with previous reports on reliability of DTI measures by using different designs and methodology. Notwithstanding the difficulties associated with region of interest–derived DTI measurements in hippocampal regions, the present approach provides estimates of distribution characteristics and precision applicable to routine assessments of DTI parameters in clinical and research context.