動的輪郭モデルを用いた超音波画像の輪郭抽出法

摘要

超音波画像はスペックルノイズが画像全体に生じるため，関心領域の境界が不鮮明となり読影には経験をrn要する．対象となる腫瘍像の輪郭を自動的に抽出できれば，超音波画像の読影の補助となりうる．本研究の目的は輝rn度勾配ベクトル場を用いた動的輪郭モデルを用い，超音波画像から関心領域の輪郭抽出を行うことである．本手法はrn画像の輝度勾配ベクトルを求め，そのベクトル場にしたがって動的輪郭モデルを関心領域の輪郭へ修正し，輪郭を抽rn出するものである．本手法を腫瘍像へ適用し輪郭の抽出を行った．%At excision operations of a breast tumor, a navigation to display three-dimensional models during the operations is demanded to grasp a position and a size of the tumor, and to decide an area of the excision. Speckle noises which are characteristic of an ultrasonic image are caused by interference of sound waves. The noises cause a low resolution of a region of interest (ROI), and those are obstacle of constructing recognizable three-dimensional image. To reconstruct a three-dimensional model from two-dimensional ultrasonic tomograms, a speckle reduction and a contour extraction of the ROI are demanded. The purpose of this study is a contour extraction of a ROI on ultrasonic images. An active contour model using a gradient vector flow was employed. The contour of a lesion area of the ultrasonic images which speckle are reduced was extracted.