How to achieve fast, accurate, and robust rigid registration between fluoroscopic X-ray and CT images

This paper presents algorithmic refinements to our method for rigid registration between preoperative CT and intraoperative X-rays and studies the influence of various setups and factors on the methods' performance. Experimental results on five different anatomical structures show that the new three-phase algorithm achieves a mean accuracy of 0.7 mm (max=1.1 mm) 95% of the time with an initial pose offset of up to 60 mm in 80 s average with three X-ray images and CT slice spacing of 2.4 mm and handles image clutter well.