安孫子 幸弘¹, 大橋 剛介¹, 伊藤 貴司², 中島 真人¹

Yukihiro ABIKO¹, Gousuke OHASI¹, Takashi ITO², Masato NAKAJIMA¹

¹慶應義塾大学理工学部電気工学科, ²アロカ株式会社

¹Department of Electrical Engineering Keio University Faculty of Science　and Technology, ²Aloka Co., Ltd.

キーワード : Boundary extraction, 4-D display, Fuzzy reasoning, M-mode, Relaxation techniques

Producing 4-D images of the heart from echocardiograms requires extraction of organ boundaries in the lumen of the heart. Because echo-image quality is not yet good enough to facilitate a fully automatic method for extracting these boundaries, we turned our attention to M-mode images, which are widely used in cardiac diagnosis. Recognition of the boundaries is relatively easy for the human eye, because boundaries appearing in M-mode images are simple, periodic waveforms that simply move up and down through time. We thus presume that it should be possible to extract organ boundaries automatically and efficiently by processing M-mode waveforms. Our method uses fuzzy reasoning to calculate the grade of each pixel's attribute, which may be blood, boundary, or soft-tissue. We then apply a de-fuzzy algorithm based on the relaxation technique to the whole image iteratively to gradually renew attribute grade. In this way, each pixel is assigned one of three attributes. To verify the effectiveness of this method, we applied it to 4-D data of human heart. This attempt to extract boundaries for use in reconstructing high quality 4-D images of the heart produced good results.