Abstract:
X-ray fluoroscopy is widely used in image-guided interventions especially in catheter-based interventions. X-ray fluoroscopy provides high temporal and spacial resolution, but it suffers from low soft tissue contrast. On the other hand, magnetic resonance imaging (MRI) offers excellent soft tissue contrast and 3D anatomical information. X-ray fused with MRI (XFM) is a system which combines strengths of both image modalities to improve the quality of imageguidance and to achieve minimally invasive interventions. In XFM, pre-operative MR images are segmented, 3D structure of target area is reconstructed from these segments, its 2D projection is overlapped on top of live images during x-ray fluoroscopy. Fusion of two images requires registration of two images which could be archived using external fiducial markers attached to skin of patient. In this approach, first markers are detected and located in both image sets, then least square minimization algorithm is applied to complete the registration. The purpose of our study is to extend the currently practiced XFM systems and to allow its translation into a practical clinical setting by making it easier to use. We developed a fully automatic registration system for XFM. This includes automatic segmentation and localization of fiducial markers in both images and finding the correspondence between two point sets, also designing a marker localization system and development of user interface for technical user. In vivo validation of our method was performed in 10 animal experiments. Results show that our method locates markers in high accuracy, finds correspondence between two point sets and completes the registration process.|Keywords: Image-Guided Intervention, Image Registration, X-Ray Fused with MRI, Correspondence Between Point Sets.
