The three dimensional (3D) reconstruction of a medical image sequence can provide intuitive morphologies of a target and help doctors to make more reliable diagnosis and give a proper treatment plan. This paper aims to reconstruct the surface of a renal corpuscle from the microscope renal biopsy image sequence. First, the contours of renal corpuscle in all slices are extracted automatically by using a context-based segmentation method with a coarse registration. Then, a new coevolutionarybased strategy is proposed to realize a fine registration. Finally, a Gauss-Seidel iteration method is introduced to achieve a non-rigid registration. Benefiting from the registrations, a smooth surface of the target can be reconstructed easily. Experimental results prove that the proposed method can effectively register the contours and give an acceptable surface for medical doctors.
|ジャーナル||IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences|
|出版ステータス||Published - 2016 12月 1|
ASJC Scopus subject areas
- コンピュータ グラフィックスおよびコンピュータ支援設計