講演抄録/キーワード |
講演名 |
2007-01-27 11:25
Automated Determination of the Liver Boundary in CT Data Based on Anatomical knowledge and Level Set Method Amir .H. Foruzan・○Reza A. Zoroofi(Univ. of Tehran)・Yoshinobu Sato・Masatoshi Hori・Takamichi Murakami・Hironobu Nakamura・Shinichi Tamurab(Osaka Univ.) |
抄録 |
(和) |
Quantitative analysis of the liver CT images is clinically important in corresponding diagnosis, treatment and surgical planning procedures. Automatic determination of the liver boundary in CT images is difficult due to similar intensity of liver with respect to its surrounding tissues, poor in-plane and inter-plane resolutions, and partial volume effect. Due to considerable changes in the location and size of the liver tissues in the abdominal duct, making robust anatomical and intensity based assumptions as well as developing robust segmentation procedures are necessary for liver boundary determination. In this research, we automatically extract required parameters according to specific features of the liver tissues. We then employ an active contour method to estimate the final boundary of the liver. After careful investigation of several donor data sets, acquired for liver transplantation, we propose a multi-step approach based on anatomical knowledge about the liver as follows. (1) We estimate the largest liver mask in the abdomen by conventional thresholding and morphological operations. (2) The mask is utilized for accurate estimation of the liver intensity range with respect to other surrounding tissues. (3) We then convert the liver gray images to color images by a heuristic color map transformation. (4) A texture analysis technique based on rotational invariant property and wavelet transform are employed to extract important features of liver regions. Based on this, liver initial boundary is estimated. (5) We employ the boundaries corresponding to initial liver regions in a level set based active contours and refine it to the final liver mask. Experiments in the presence of 4000 CT images (acquired from 20 donors) confirm the success of the proposed techniques. |
(英) |
Quantitative analysis of the liver CT images is clinically important in corresponding diagnosis, treatment and surgical planning procedures. Automatic determination of the liver boundary in CT images is difficult due to similar intensity of liver with respect to its surrounding tissues, poor in-plane and inter-plane resolutions, and partial volume effect. Due to considerable changes in the location and size of the liver tissues in the abdominal duct, making robust anatomical and intensity based assumptions as well as developing robust segmentation procedures are necessary for liver boundary determination. In this research, we automatically extract required parameters according to specific features of the liver tissues. We then employ an active contour method to estimate the final boundary of the liver. After careful investigation of several donor data sets, acquired for liver transplantation, we propose a multi-step approach based on anatomical knowledge about the liver as follows. (1) We estimate the largest liver mask in the abdomen by conventional thresholding and morphological operations. (2) The mask is utilized for accurate estimation of the liver intensity range with respect to other surrounding tissues. (3) We then convert the liver gray images to color images by a heuristic color map transformation. (4) A texture analysis technique based on rotational invariant property and wavelet transform are employed to extract important features of liver regions. Based on this, liver initial boundary is estimated. (5) We employ the boundaries corresponding to initial liver regions in a level set based active contours and refine it to the final liver mask. Experiments in the presence of 4000 CT images (acquired from 20 donors) confirm the success of the proposed techniques. |
キーワード |
(和) |
Liver Segmentation / Texture Based Segmentation / Wavelet / Active contours / Level Set / / / |
(英) |
Liver Segmentation / Texture Based Segmentation / Wavelet / Active contours / Level Set / / / |
文献情報 |
信学技報, vol. 106, no. 510, MI2006-179, pp. 83-86, 2007年1月. |
資料番号 |
MI2006-179 |
発行日 |
2007-01-20 (MI) |
ISSN |
Print edition: ISSN 0913-5685 |
PDFダウンロード |
|
|