Radial-Biased Tracking Method to Assess Tissue Displacement in Intravascular Ultrasound Sequences: A Phantom Framework Evaluation.

Objectives: We created and evaluated a pixel-tracking method capable of accurately identify the displacement of tissue in intravascular ultrasound (IVUS) images.
Methods: Our proposed pixel-tracking method assessed the horizontal and vertical displacement of tissue from a numerical phantom of IVUS sequences. The proposed tracking method is based on a block-matching framework, comparing 2 distinct frames within a selected region by normalized cross-correlation. Our method, specialized for IVUS applications, reduced the tracking area by implementing a limiting radius and a radial bias during the search.
Results: The method was evaluated by using 54 numerical phantom image sequences from 9 distinct arterial models, resulting in different arteries with atherosclerotic plaques under a range of pressures. The ground truth reference coordinates of the tracked tissue were extracted from each numerical phantom sequence. Our results were compared to 8 other methods present in the literature. The mean absolute tracking errors ± SD for our method were 15.56 ± 19.46 and 13.04 ± 13.82 μm for the horizontal and vertical directions, respectively, between 2 subsequent frames, and 162.58 ± 305.93 and 102.22 ± 130.61 μm from lower to higher pressures in the range of 6 frames (n = 42,036).
Conclusions: Our application-specific pixel-tracking method showed promising results and no statistically significant tracking error (P = .954), comparable to state-of-the-art methods present in the literature. Application-specific tracking methods have advantages over general methods by turning tissue-specific behavior into a directional bias in the tracking algorithm.
(© 2019 by the American Institute of Ultrasound in Medicine.)