Quantification of rotational tissue motion in IVUS with optical flow algorithm

A novel method for estimating tissue motion in two-dimensional intravascular ultrasound image sequences of coronary arteries is proposed. It is based on the classical algorithm by Lukas-Kanade for optical flow. The optical flow vector field describes the transformation between consecutive frames and quantifies the amount of apparent tissue displacement. It is estimated for each point of the vessel wall in the least-square sense over a sliding window. However, instead of assuming the motion constancy, a locally affine motion model is adopted. It effectively captures and appropriately describes a complex deformation pattern of coronary vessel walls under variable physiological conditions (i.e. heart contraction, pulsatile blood pressure). Another refinement contains a novel feature-based weighting scheme, which prioritizes the contribution of each spatial neighbor depending on its traceability. Finally, the scale-space embedding of the classical optical flow equation is derived. The inaccuracy in the computed flow field with respect to the true solution for different scales (Gaussian blurring) and weighting schemes was estimated. This systematic approach, carried out on the artificially rotated images of a tissue-mimicking phantom, yields the optimal choice for the algorithm's parameters. The validation of the proposed method was done on het-erogenous tissue-mimicking phantoms. The experimental results showed robust and reliable performance of the algorithm up to 4.5°, which is in the plausible range of circumferential vessel wall displacement of coronary arteries.

• View at Publisher
• View at Scopus