Depth-resolved phase velocity estimation in layered tissue based on an efficient additive attention network with surface acoustic wave – optical coherence elastography

TY - JOUR

T1 - Depth-resolved phase velocity estimation in layered tissue based on an efficient additive attention network with surface acoustic wave – optical coherence elastography

AU - Zhang, Guangyu

AU - Liao, Jinpeng

AU - Feng, Zhengshuyi

AU - van Bocxlaer, Katrien

AU - Layton, Alison M.

AU - Li, Chunhui

AU - Huang, Zhihong

N1 - Publisher Copyright: © 2026 Optica Publishing Group (formerly OSA). All rights reserved.

PY - 2026/5/1

Y1 - 2026/5/1

N2 - Optical coherence elastography (OCE) is a non-invasive imaging technique used to quantify tissue stiffness and to assist in the diagnosis and assessment of disease. A major limitation of conventional OCE approaches is that phase velocity estimation requires transformation from the spatial–temporal domain to the frequency–wavenumber domain, a process that is computationally inefficient and may introduce errors due to assumptions regarding tissue properties. We propose a unified framework for depth-resolved phase velocity estimation that combines spectral analysis of complex-valued signals with a deep learning inversion network. The effectiveness of the framework is validated using homogeneous agar phantoms, while layered agar phantoms and in vivo human skin are analyzed by depth-dependent phase velocity gradients. The proposed phase velocity estimation network (PVNet) achieved a mean absolute error (MAE) of 0.123 ± 0.024 m/s in agar models and 0.145 ± 0.114 m/s in human skin, compared with ground truth measurements. This study presents a deep learning approach for segmenting depth-resolved bi-layers in OCE, offering significant potential for the clinical identification of sub-surface lesions and abnormalities.

AB - Optical coherence elastography (OCE) is a non-invasive imaging technique used to quantify tissue stiffness and to assist in the diagnosis and assessment of disease. A major limitation of conventional OCE approaches is that phase velocity estimation requires transformation from the spatial–temporal domain to the frequency–wavenumber domain, a process that is computationally inefficient and may introduce errors due to assumptions regarding tissue properties. We propose a unified framework for depth-resolved phase velocity estimation that combines spectral analysis of complex-valued signals with a deep learning inversion network. The effectiveness of the framework is validated using homogeneous agar phantoms, while layered agar phantoms and in vivo human skin are analyzed by depth-dependent phase velocity gradients. The proposed phase velocity estimation network (PVNet) achieved a mean absolute error (MAE) of 0.123 ± 0.024 m/s in agar models and 0.145 ± 0.114 m/s in human skin, compared with ground truth measurements. This study presents a deep learning approach for segmenting depth-resolved bi-layers in OCE, offering significant potential for the clinical identification of sub-surface lesions and abnormalities.

UR - https://www.scopus.com/pages/publications/105037861579

U2 - 10.1364/BOE.593027

DO - 10.1364/BOE.593027

M3 - Article

C2 - 42145698

AN - SCOPUS:105037861579

SN - 2156-7085

VL - 17

SP - 2533

EP - 2548

JO - Biomedical Optics Express

JF - Biomedical Optics Express

IS - 5

ER -