Regression-based algorithm for bulk motion subtraction in optical coherence tomography angiography

Acner Camino, Jia Yali, Gangjun Liu, Jie Wang, David Huang

Research output: Contribution to journalArticle

12 Citations (Scopus)

Abstract

We developed an algorithm to remove decorrelation noise due to bulk motion in optical coherence tomography angiography (OCTA) of the posterior eye. In this algorithm, OCTA B-frames were divided into segments within which the bulk motion velocity could be assumed to be constant. This velocity was recovered using linear regression of decorrelation versus the logarithm of reflectance in axial lines (A-lines) identified as bulk tissue by percentile analysis. The fitting parameters were used to calculate a reflectance-adjusted upper bound threshold for bulk motion decorrelation. Below this threshold, voxels are identified as non-flow tissue, their flow values are set to zeros. Above this threshold, the voxels are identified as flow voxels and bulk motion velocity is subtracted from each using a nonlinear decorrelation-velocity relationship previously established in laboratory flow phantoms. Compared to the simpler median-subtraction method, the regression-based bulk motion subtraction improved angiogram signal-to-noise ratio, contrast, vessel density repeatability, and bulk motion noise cleanup in the foveal avascular zone, while preserving the connectivity of the vascular networks in the angiogram.

Original languageEnglish (US)
Article number#290746
JournalBiomedical Optics Express
Volume8
Issue number6
DOIs
StatePublished - Jun 1 2017

Fingerprint

angiography
Optical Coherence Tomography
subtraction
regression analysis
Angiography
tomography
thresholds
Noise
reflectance
Signal-To-Noise Ratio
logarithms
preserving
vessels
Blood Vessels
Linear Models
signal to noise ratios

Keywords

  • Image enhancement
  • Motion detection
  • Ophthalmology
  • Optical coherence tomography

ASJC Scopus subject areas

  • Biotechnology
  • Atomic and Molecular Physics, and Optics

Cite this

Regression-based algorithm for bulk motion subtraction in optical coherence tomography angiography. / Camino, Acner; Yali, Jia; Liu, Gangjun; Wang, Jie; Huang, David.

In: Biomedical Optics Express, Vol. 8, No. 6, #290746, 01.06.2017.

Research output: Contribution to journalArticle

@article{4255ad6050cd4557a333bf2eb3c3c830,
title = "Regression-based algorithm for bulk motion subtraction in optical coherence tomography angiography",
abstract = "We developed an algorithm to remove decorrelation noise due to bulk motion in optical coherence tomography angiography (OCTA) of the posterior eye. In this algorithm, OCTA B-frames were divided into segments within which the bulk motion velocity could be assumed to be constant. This velocity was recovered using linear regression of decorrelation versus the logarithm of reflectance in axial lines (A-lines) identified as bulk tissue by percentile analysis. The fitting parameters were used to calculate a reflectance-adjusted upper bound threshold for bulk motion decorrelation. Below this threshold, voxels are identified as non-flow tissue, their flow values are set to zeros. Above this threshold, the voxels are identified as flow voxels and bulk motion velocity is subtracted from each using a nonlinear decorrelation-velocity relationship previously established in laboratory flow phantoms. Compared to the simpler median-subtraction method, the regression-based bulk motion subtraction improved angiogram signal-to-noise ratio, contrast, vessel density repeatability, and bulk motion noise cleanup in the foveal avascular zone, while preserving the connectivity of the vascular networks in the angiogram.",
keywords = "Image enhancement, Motion detection, Ophthalmology, Optical coherence tomography",
author = "Acner Camino and Jia Yali and Gangjun Liu and Jie Wang and David Huang",
year = "2017",
month = "6",
day = "1",
doi = "10.1364/BOE.8.003053",
language = "English (US)",
volume = "8",
journal = "Biomedical Optics Express",
issn = "2156-7085",
publisher = "The Optical Society",
number = "6",

}

TY - JOUR

T1 - Regression-based algorithm for bulk motion subtraction in optical coherence tomography angiography

AU - Camino, Acner

AU - Yali, Jia

AU - Liu, Gangjun

AU - Wang, Jie

AU - Huang, David

PY - 2017/6/1

Y1 - 2017/6/1

N2 - We developed an algorithm to remove decorrelation noise due to bulk motion in optical coherence tomography angiography (OCTA) of the posterior eye. In this algorithm, OCTA B-frames were divided into segments within which the bulk motion velocity could be assumed to be constant. This velocity was recovered using linear regression of decorrelation versus the logarithm of reflectance in axial lines (A-lines) identified as bulk tissue by percentile analysis. The fitting parameters were used to calculate a reflectance-adjusted upper bound threshold for bulk motion decorrelation. Below this threshold, voxels are identified as non-flow tissue, their flow values are set to zeros. Above this threshold, the voxels are identified as flow voxels and bulk motion velocity is subtracted from each using a nonlinear decorrelation-velocity relationship previously established in laboratory flow phantoms. Compared to the simpler median-subtraction method, the regression-based bulk motion subtraction improved angiogram signal-to-noise ratio, contrast, vessel density repeatability, and bulk motion noise cleanup in the foveal avascular zone, while preserving the connectivity of the vascular networks in the angiogram.

AB - We developed an algorithm to remove decorrelation noise due to bulk motion in optical coherence tomography angiography (OCTA) of the posterior eye. In this algorithm, OCTA B-frames were divided into segments within which the bulk motion velocity could be assumed to be constant. This velocity was recovered using linear regression of decorrelation versus the logarithm of reflectance in axial lines (A-lines) identified as bulk tissue by percentile analysis. The fitting parameters were used to calculate a reflectance-adjusted upper bound threshold for bulk motion decorrelation. Below this threshold, voxels are identified as non-flow tissue, their flow values are set to zeros. Above this threshold, the voxels are identified as flow voxels and bulk motion velocity is subtracted from each using a nonlinear decorrelation-velocity relationship previously established in laboratory flow phantoms. Compared to the simpler median-subtraction method, the regression-based bulk motion subtraction improved angiogram signal-to-noise ratio, contrast, vessel density repeatability, and bulk motion noise cleanup in the foveal avascular zone, while preserving the connectivity of the vascular networks in the angiogram.

KW - Image enhancement

KW - Motion detection

KW - Ophthalmology

KW - Optical coherence tomography

UR - http://www.scopus.com/inward/record.url?scp=85020220870&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85020220870&partnerID=8YFLogxK

U2 - 10.1364/BOE.8.003053

DO - 10.1364/BOE.8.003053

M3 - Article

AN - SCOPUS:85020220870

VL - 8

JO - Biomedical Optics Express

JF - Biomedical Optics Express

SN - 2156-7085

IS - 6

M1 - #290746

ER -