Optimization of b-value distribution for biexponential diffusion-weighted MR imaging of normal prostate

J Magn Reson Imaging. 2014 May;39(5):1213-22. doi: 10.1002/jmri.24271. Epub 2013 Oct 11.

Abstract

Purpose: To determine the optimal b-value distribution for biexponential diffusion-weighted imaging (DWI) of normal prostate using both a computer modeling approach and in vivo measurements.

Materials and methods: Optimal b-value distributions for the fit of three parameters (fast diffusion Df, slow diffusion Ds, and fraction of fast diffusion f) were determined using Monte-Carlo simulations. The optimal b-value distribution was calculated using four individual optimization methods. Eight healthy volunteers underwent four repeated 3 Tesla prostate DWI scans using both 16 equally distributed b-values and an optimized b-value distribution obtained from the simulations. The b-value distributions were compared in terms of measurement reliability and repeatability using Shrout-Fleiss analysis.

Results: Using low noise levels, the optimal b-value distribution formed three separate clusters at low (0-400 s/mm2), mid-range (650-1200 s/mm2), and high b-values (1700-2000 s/mm2). Higher noise levels resulted into less pronounced clustering of b-values. The clustered optimized b-value distribution demonstrated better measurement reliability and repeatability in Shrout-Fleiss analysis compared with 16 equally distributed b-values.

Conclusion: The optimal b-value distribution was found to be a clustered distribution with b-values concentrated in the low, mid, and high ranges and was shown to improve the estimation quality of biexponential DWI parameters of in vivo experiments.

Keywords: MRI; Monte-Carlo simulations; Shrout-Fleiss analysis; diffusion-weighted imaging; normal prostate.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Adult
  • Algorithms*
  • Computer Simulation
  • Data Interpretation, Statistical
  • Humans
  • Image Enhancement / methods
  • Image Interpretation, Computer-Assisted / methods*
  • Male
  • Middle Aged
  • Models, Biological*
  • Models, Statistical
  • Prostate / anatomy & histology*
  • Reference Values
  • Reproducibility of Results
  • Sensitivity and Specificity
  • Statistical Distributions