Abstract
Spatio-temporal registration of baseline and follow-up intravascular ultrasound (IVUS) pullbacks is of paramount importance in studying the progression/regression of coronary artery disease. Automating these two tasks has the potential to increase productivity when studying large patient populations.
Current automated methods are often designed for only one of the two tasks - spatial or temporal. In this paper, we propose an integrated framework which combines the two tasks and employs side-branches to constrain the IVUS pullback registration tasks.
For temporal registration, canonical time warping technique optimizes extracted features and weighs cumulative distances. For spatial registration, the search range of cross-correlation based method is constrained by utilizing the angular differences between side-branches.
Pilot validation is currently available for ten pairs of IVUS pullback sub-sequences. Results show average spatial and temporal registration errors of 0.49 mm +/- 0.51 mm and 5.56 degrees +/- 3.35 degrees, respectively, a notable improvement over our previous approach (p < 0.001) in temporal registration.
Our method has the potential to improve spatial and temporal correspondence in studies of atherosclerotic vascular disease development using IVUS.