PT  - JOURNAL ARTICLE
AU  - Lee, Hakjae
AU  - Lee, Kisung
AU  - Kim, Kyeong Min
AU  - Chun, Gi-Jeong
AU  - Joo, Sung-Kwan
TI  - New strategy of full automatic 3D registration for abdominal PET/CT and MRI images
DP  - 2010 May 01
TA  - Journal of Nuclear Medicine
PG  - 1307--1307
VI  - 51
IP  - supplement 2
4099  - http://jnm.snmjournals.org/content/51/supplement_2/1307.short
4100  - http://jnm.snmjournals.org/content/51/supplement_2/1307.full
SO  - J Nucl Med2010 May 01; 51
AB  - 1307 Objectives Accurate registration of abdominal images acquired PET and MRI is essential for the relevant diagnosis using whole-body image, as much as brain studies. Despite of many studies for neuroimages, it has been few studies of image registration for abdominal PET and MRI images. In this study, we developed a full automatic 3D image registration algorithm for abdominal PET and MRI images. Methods The proposed algorithm consists of two steps; global registration and feature-based adjustment. Global registration was achieved with affine transformation which was accelerated by CUDA technology, normalized mutual information(NMI) and particle swarm optimization(PSO). In the feature-based adjustment, we extracted the 2D feature planes from a CT slice and multiple MR slices at similar locations to the CT slice by independent component analysis(ICA). The parameters for free-form deformation(FFD) were extracted by measuring similarity between the extracted ICA plane and MR slices. This process was separately conducted with 3D planes then each control point has a different set of FFD parameters. Finally, MR volume was transformed to match the PET/CT volume. Results The proposed method showed that, lesions in PET images successfully matched with those in MR images. In global registration, PSO achieved significant reduction of processing time and excellent reproducibility. Misaligned parts in the global registration step were successfully corrected by using ICA and FFD so that the ROI in MR images are more precisely matched with that in PET images. Conclusions The proposed method effectively matches two ROI’s in PET/CT images and standalone MR images. The reduction of processing time and reproducibility of global registration algorithm showed its effectiveness. The ICA extracted salient feature information and improves the registration performance significantly. The proposed algorithm was also successfully employed in PET/CT and CT registration