The application of the local histograms of apparent difusion coefficient in differentiation of brain astrocytomas

Abstract

Background/Aim. Microstructural diversity of brain astrocytomas makes their diagnostics and differentiation by using the diffusion weighted imaging (DWI) difficult. In this study we used the histogram-based positioning of regions of interests on the apparent diffusion coefficient (ADC) maps in order to restrict the determination of diffusion parameters to regions of interest (ROI) corresponding to maximum cellularity. Success of ADC standard deviation (∆ADC) and kurtosis (K) in differentiation of brain astrocytomas was evaluated. Methods. The thirtyone patients (16 women and 15 men, median age 37 years, age range 6–72 years) with suspected supratentorial astrocytomas were included in the retrospective study. The magnetic resonance imaging (MRI) examinations were performed using the 1.5 T MR system (Avanto; Siemens, Erlangen, Germany) and 8-channel phased array head coil. The DWI images were acquired in three orthogonal directions for the b-values 0, 500 and 1000 s mm-2. The histogram calculations and determination of diffusion parameters were performed using the MIPAV software package and the statistical analysis was done in the Openstat software. Results. The ADC values enabled differentiation of diffuse astrocytomas (DA) from a high-grade astrocytoma (HGA), but not between the classes of HGA. In addition, the ∆ADC value provided discrimination between the anaplastic astrocytoma (AA) and glioblastoma multiforme (GBM) with 100% of sensitivity and 89% of specificity. The kurtosis value can also differentiate between the grades AA and GBM although with the lower sensitivity and specificity. Conclusion. The histogram analysis of tumor region on the ADC maps can provide a guidance for an appropriate choice of the ROIs. The parameters which characterize diffusion of such defined ROIs, as well as their combination can improve differentiation of brain astrocytomas.