大橋 剛介¹, 川崎 育代¹, 秋山 いわき², 中島 真人¹

Gosuke OHASHI¹, Ikuyo KAWASAKI¹, Iwaki AKIYAMA², Masato NAKAJIMA¹

¹慶應義塾大学理工学部電気工学科中島研究室, ²湘南工科大学工学部

¹Department of Electrical Engineering, Keio University, ²Department of Electrical Engineering, Shonan Institute of Technology

キーワード : Boundary extraction, Echographic image, LOG filter, Prostate, Transabdominal data

We report an attempt to develop a highly accurate algorithm that can automatically extract the image of the prostate from echographic images. The method comprises three steps: first, a series of echographic images obtained by transabdominal scan provide three dimensional data on the prostate; second, a Laplacian of Gaussian (LOG) filter is applied to the echographic images to detect the boundary within the prostate region; last, the prostate region is extracted using prior knowledge. Because two-dimensional LOG filtering was not accurate enough to detect boundaries, we extended the LOG filter to three dimensions. Closed-curve surfaces found with boundary detection correspond to the target region but include incomplete and false surfaces as well. Prior knowledge is used to extract the true prostate region from these surfaces. This method extracted most of the prostate region automatically in some clinical cases.