3D curve constrained deformable registration using a neuro-fuzzy transformation model

Image registration of abdominal organs and soft tissues is considered daunting due to large organ shift and tissue deformation caused by patient motion, respiration, etc. In this study, we propose a novel neuro-fuzzy deformable registration technique that is constrained by 3D curves of vessel centerlines and point marks while minimizing strain energy. We present an analytical global optimal solution in the case when 3D curves, strain energy and point marks are considered, which will provide fast and robust deformable match for internal structures such as blood vessels, and significantly reduce the chance to get trapped in local minima. We have demonstrated the effectiveness of our deformable technique in registering liver MR images. Validation shows a target registration error of 1.98 mm and an average centerline distance error of 1.65 mm. This technique has the potential to significantly improve registration capability and the quality of intra-operative image guidance.