Brain tumor DWIs: Comparing the results of manual and computer-based evaluation

Abstract

The article is focused on the segmentation of pathology in brain tissue derived from diffusion weighted images. The authors describe two approaches of segmentation. The first approach is based on support vector machine classification and the second on manual segmentation. Segmentation was performed on four types of images that have been derived from calculation of ADC (apparent diffusion coefficient) map (Trace, Fraction Anisotropy, Volume Ratio, and Mean of All B0s). Segmented data was analyzed statistically to compare two researched approaches of segmentation. Additionally, the article investigates the most valuable type of processed images for detection and classification of pathologic tissue — tumor.