[Paper Review] Spatio-temporal normalized cross-correlation for estimation of the displacement field in ultrasound elastography

Morteza Mirzaei, Amir Asif+2 more|arXiv (Cornell University)|Apr 15, 2018

Ultrasound Imaging and ElastographyMedicine36 references4 citations

TL;DR

This paper proposes Spatio-Temporal Normalized Cross-Correlation (STNCC), a novel displacement estimation method in ultrasound elastography that extends traditional spatial windowing to include temporal coherence across multiple frames. By leveraging spatio-temporal windows, STNCC improves robustness to noise and signal decorrelation, achieving up to 227% higher SNR and 67% higher CNR than conventional NCC in in-vivo liver data.

ABSTRACT

This paper introduces a novel technique to estimate tissue displacement in quasi-static elastography. A major challenge in elastography is estimation of displacement (also referred to time-delay estimation) between pre-compressed and post-compressed ultrasound data. Maximizing normalized cross correlation (NCC) of ultrasound radio-frequency (RF) data of the pre- and post-compressed images is a popular technique for strain estimation due to its simplicity and computational efficiency. Several papers have been published to increase the accuracy and quality of displacement estimation based on NCC. All of these methods use spatial windows to estimate NCC, wherein displacement magnitude is assumed to be constant within each window. In this work, we extend this assumption along the temporal domain to exploit neighboring samples in both spatial and temporal directions. This is important since traditional and ultrafast ultrasound machines are, respectively, capable of imaging at more than 30 frame per second (fps) and 1000 fps. We call our method spatial temporal normalized cross correlation (STNCC) and show that it substantially outperforms NCC using simulation, phantom and in-vivo experiments.

Motivation & Objective

• To address the limitation of spatial-only windowing in normalized cross-correlation (NCC) for time-delay estimation in ultrasound elastography.
• To improve displacement estimation accuracy and noise robustness by exploiting temporal continuity in high-frame-rate ultrasound sequences.
• To extend the standard NCC framework to include temporal neighbors, enabling spatio-temporal window matching.
• To validate the method across simulation, phantom, and in-vivo datasets, demonstrating superior performance over conventional NCC.

Proposed method

• The method introduces spatio-temporal windows that span both spatial and temporal dimensions, capturing displacement consistency across multiple frames.
• It formulates a normalized cross-correlation (NCC) metric over 3D windows (x, y, t), enhancing estimation stability by aggregating information across time.
• The algorithm matches corresponding spatio-temporal windows between pre-compressed and post-compressed ultrasound RF data to estimate displacement fields.
• It assumes constant displacement within each 3D window, extending the classical spatial-only NCC assumption into the temporal domain.
• The method is applied to radio-frequency (RF) ultrasound data from two image sequences: pre- and post-compression.
• The final displacement field is computed by maximizing the STNCC metric across all possible time-delay shifts within the search window.

Experimental results

Research questions

• RQ1 Can incorporating temporal coherence into the NCC framework improve displacement estimation accuracy in ultrasound elastography?
• RQ2 How does STNCC perform under varying levels of signal decorrelation and noise compared to conventional NCC?
• RQ3 Does the inclusion of multiple temporal frames reduce estimation variance and improve robustness in real-world imaging scenarios?
• RQ4 How does the performance of STNCC vary with different window overlap ratios in practical settings?
• RQ5 To what extent does STNCC enhance SNR and CNR in in-vivo and phantom data compared to standard NCC?

Key findings

• STNCC improved SNR by 146.66% at 70% window overlap and by 226.82% at 30% overlap in back muscle phantom data compared to NCC.
• In in-vivo liver data, STNCC increased SNR by 71.06% and CNR by 67.15% compared to NCC.
• STNCC demonstrated superior robustness to window overlap variations, maintaining high-quality displacement fields even at low overlaps.
• Visual and quantitative analysis confirmed that STNCC produces strain images with significantly less noise and higher contrast.
• The method effectively reduces estimation variance by leveraging temporal redundancy in high-frame-rate ultrasound sequences.
• The results validate that extending the NCC window into the temporal domain significantly enhances displacement estimation performance across diverse imaging conditions.