Performance assessment of wave speed estimators based on spatial-frequency analysis in quasi-harmonic optical coherence elastography

Wave-speed (WS) estimators based on spatial-frequency analysis, such as the Continuous Wavelet Transform (CWT) and the Adaptive Superlet Transform (ASLT), have demonstrated their potential for Ultrasound Crawling Wave Sonoelastography. However, they have not yet been validated for Optical Coherence Elastography (OCE). To address this, a quasi-harmonic mechanical excitation OCE (QME-OCE) system was used to explore the elasticity of a two-sided gelatin phantom (soft side: 6% - hard side: 9%) by estimating WS maps. We hypothesize that ASLT and CWT algorithms can enhance the estimation of phantom WS distributions obtained with the Phase Derivative (PD) method—even when fewer particle velocity frames were used. QME-OCE involved a swept source phase-sensitive Optical Coherence Tomography system and a piezoelectric actuator for contact excitation vibrating at 700, 800, and 900 Hz. We indirectly measured shear wave speed (SWS) from Rayleigh waves propagation and evaluated the performance of the algorithms using mechanical measurements (SWS6%=3.93±0.47m/s and SWS9%=4.84±0.88m/s). Our findings demonstrated that ASLT and CWT outperform PD in terms of precision error for 6% and 9% (PD6%=19.05±2.07%, CWT6%=11.07±1.75%, ASLT6%=7.99±1.85%; and PD9%=18.74±1.45%, CWT9%=7.46±3.74%, ASLT9%=3.57±1.13%); and accuracy error for 6% side (PD6%=18.58±6.46%, CWT6%=16.05±3.94%, ASLT6%=15.69±3.24%; and PD9%=7.49±4.83%, CWT9%=8.74±3.88%, ASLT9%=11.46±9.06%). Additionally, the proposed algorithms surpassed PD in terms of contrast-to-noise ratio PD=1.57±0.09, CWT=2.70±1.03, ASLT=2.99±0.87), but PD is superior in terms of lateral resolution (PD=0.68±0.32mm, CWT=1.76±0.11mm, ASLT=3.25±0.66mm). This indicates that ASLT and CWT provide improved differentiation of stiffer regions, while PD allows for faster detection of stiffness changes. Furthermore, time-downsampling analysis demonstrated that acquisition sequences can be substantially shortened for CWT and ASLT without adversely affecting SWS estimation.