An Image Registration Based Technique for Noninvasive Vascular Elastography

TL;DR

This paper introduces a non-rigid image registration technique for non-invasive vascular elastography that improves displacement measurement accuracy under larger compressions, outperforming traditional cross correlation methods.

Contribution

The study proposes a novel non-rigid image registration approach that reduces displacement measurement errors in vascular elastography, especially at higher compressions, and enhances image quality using synthetic aperture imaging.

Findings

Proposed method reduces displacement error with increased compression.

Achieved lower relative RMSE (4.5%) compared to cross correlation (6%).

Outperforms conventional methods by 25% in error reduction.

Abstract

Non-invasive vascular elastography is an emerging technique in vascular tissue imaging. During the past decades, several techniques have been suggested to estimate the tissue elasticity by measuring the displacement of the Carotid vessel wall. Cross correlation-based methods are the most prevalent approaches to measure the strain exerted in the wall vessel by the blood pressure. In the case of a low pressure, the displacement is too small to be apparent in ultrasound imaging, especially in the regions far from the center of the vessel, causing a high error of displacement measurement. On the other hand, increasing the compression leads to a relatively large displacement in the regions near the center, which reduces the performance of the cross correlation-based methods. In this study, a non-rigid image registration-based technique is proposed to measure the tissue displacement for a relatively large compression. The results show that the error of the displacement measurement obtained by the proposed method is reduced by increasing the amount of compression while the error of the cross correlationbased method rises for a relatively large compression. We also used the synthetic aperture imaging method, benefiting the directivity diagram, to improve the image quality, especially in the superficial regions. The best relative root-mean-square error (RMSE) of the proposed method and the adaptive cross correlation method were 4.5% and 6%, respectively. Consequently, the proposed algorithm outperforms the conventional method and reduces the relative RMSE by 25%.