Abstract
Depth-resolved optical attenuation coefficient is a valuable tissue parameter that complements the intensity-based structural information in optical coherent tomography (OCT) imaging. Herein we systematically analyzed the under- and over-estimation bias of existing depth-resolved methods when applied to real biological tissues, and then proposed a new algorithm that remedies these issues and accommodates general OCT data that contain incomplete decay and noise floor, thereby affording consistent estimation accuracy for practical biological samples of different scattering properties. Compared with other algorithms, our method demonstrates remarkably improved estimation accuracy and numerical robustness, as validated via numerical simulations and on experimental OCT data obtained from both silicone-TiO2 phantoms and human ventral tongue leukoplakia samples.
Highlights
In direct analogy to ultrasound imaging, optical coherence tomography (OCT) acquires crosssectional images of biological tissues by measuring back-scattered light from different depths [1,2]
As a simplified model, denoting the depth at which such transition occurs by F (e.g., 5 dB above the noise floor intensity), we find that the attenuation coefficient given by Eq (3) for shallower depths z
Digital phantom simulation To validate the accuracy of our algorithm, we started with five single-layer, homogenous numerical phantoms with distinct attenuation coefficients
Summary
In direct analogy to ultrasound imaging, optical coherence tomography (OCT) acquires crosssectional images of biological tissues by measuring back-scattered light from different depths [1,2]. Conventional approach to quantifying tissue attenuation involves typically modeling the tissue as a homogeneous slab and fitting part of OCT A-line signal to a single-exponential decay curve [10,11,12,13,14]. Such simplified treatment smears the depth-resolved ability of OCT and disregards the depth-dependent attenuation coefficient of practical heterogeneous biological tissues. The model in discrete domain is transcribed below: μ[n] ≈
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have