専門医の診断支援を目的としたＭＲＩ脳画像データセットにおける中心近傍画像選択法に関する基礎的検討

Abstract

Multiple MR brain images are acquired at once. To begin with, the medical specialists examine a MR brain image, which shows temporal lobes, a frontal lobe, and some important regions for diagnosis, located near the cerebral center. Automatic selection of the main image from MR brain image set is a useful diagnosis imaging support for medical specialists. This paper proposes a method for extracting a ventricle area in order to select a main image. The proposed method works three steps. First, the process using a mode method that extracts sliced areas from original MR brain images. Second, the Otsu's method performs a binary process.Third, a ventricle area is a unique shape and we therefore defined two rectangles: the frontal horn and the posterior horn of a lateral ventricle region. The experimental results revealed that the proposed algorithm was able to accurately extract the ventricle area from the image set.