• 목록
• 게시판 리스트 옵션
  • 검색

Byung Jin Kim, Yong Hoon Jeong, Moon Seon Kim1, Jeom Jin Lim, Hong Jae Lee

Department of Nuclear Medicine, Seoul National University Hospital, Seoul, Korea

1Department of Radiology, Seoul National University Hospital Seoul, Korea

Purpose  :  We  expect  that  PET-MR  image  registration  could  improve diagnostic accuracy by integrating functional and anatomical images. The aim of the present study was to determine the accuracy of PET-MR image registration using Flanged Jaszczak ECT Phantom (FJECTP). Materials and Methods : We acquired PET and MR images of FJECTP with changing 6 degrees of freedom which show a motion of a rigid body in three dimensional space. The attenuation-corrected PET and MRI volume datasets were registered  by  mutual  information  algorithm.  The   angle  and  distance differences between the reference and floating images were measured before and after fusion by TrueD (Siemens Medical Solutions USA, Inc.). Result : Images were correctly matched when the distances between centers of PET and MR images were 8 cm or less in X,Y,Z directions. However, truncation artifact happened when the distances were more than 8 cm, even though image registration was achieved. Images were correctly matched when the angle differences were less than 45 degrees before fusion in X, Y axes. The limits of angle difference in Z axis for successful image registration were 15 degrees without rotation in X, Y axes and 10 degrees with 45 degrees rotation in X and Y axes, respectively. Conclusion : Images acquired in the same region by PET/MRI equipment were correctly matched using image registration based on mutual information, even though six degrees of freedom were different. Specifically, we suggest that differences of angle in Z-axis can be an important factor to achieve exact image registration because there is a limitation angle of Z-axis which can perform image registration.