Parallel deep neural networks for endoscopic oct image segmentation

Dawei Li, Jimin Wu, Yufan He, Xinwen Yao, Wu Yuan, Defu Chen, Hyeon Cheol Park, Shaoyong Yu, Jerry Ladd Prince, Xingde Li

Research output: Contribution to journalArticle

Abstract

We report parallel-trained deep neural networks for automated endoscopic OCT image segmentation feasible even with a limited training data set. These U-Net-based deep neural networks were trained using a modified dice loss function and manual segmentations of ultrahigh-resolution cross-sectional images collected by an 800 nm OCT endoscopic system. The method was tested on in vivo guinea pig esophagus images. Results showed its robust layer segmentation capability with a boundary error of 1.4 µm insensitive to lay topology disorders. To further illustrate its clinical potential, the method was applied to differentiating in vivo OCT esophagus images from an eosinophilic esophagitis (EOE) model and its control group, and the results clearly demonstrated quantitative changes in the top esophageal layers’ thickness in the EOE model.

Original languageEnglish (US)
Article number#349139
Pages (from-to)1126-1135
Number of pages10
JournalBiomedical Optics Express
Volume10
Issue number3
DOIs
StatePublished - Mar 1 2019

Fingerprint

Eosinophilic Esophagitis
Esophagus
esophagus
Guinea Pigs
guinea pigs
Control Groups
education
topology
disorders
Datasets

ASJC Scopus subject areas

  • Biotechnology
  • Atomic and Molecular Physics, and Optics

Cite this

Parallel deep neural networks for endoscopic oct image segmentation. / Li, Dawei; Wu, Jimin; He, Yufan; Yao, Xinwen; Yuan, Wu; Chen, Defu; Park, Hyeon Cheol; Yu, Shaoyong; Prince, Jerry Ladd; Li, Xingde.

In: Biomedical Optics Express, Vol. 10, No. 3, #349139, 01.03.2019, p. 1126-1135.

Research output: Contribution to journalArticle

Li, Dawei ; Wu, Jimin ; He, Yufan ; Yao, Xinwen ; Yuan, Wu ; Chen, Defu ; Park, Hyeon Cheol ; Yu, Shaoyong ; Prince, Jerry Ladd ; Li, Xingde. / Parallel deep neural networks for endoscopic oct image segmentation. In: Biomedical Optics Express. 2019 ; Vol. 10, No. 3. pp. 1126-1135.
@article{86db8f918f5b4a20892620a8a7071835,
title = "Parallel deep neural networks for endoscopic oct image segmentation",
abstract = "We report parallel-trained deep neural networks for automated endoscopic OCT image segmentation feasible even with a limited training data set. These U-Net-based deep neural networks were trained using a modified dice loss function and manual segmentations of ultrahigh-resolution cross-sectional images collected by an 800 nm OCT endoscopic system. The method was tested on in vivo guinea pig esophagus images. Results showed its robust layer segmentation capability with a boundary error of 1.4 µm insensitive to lay topology disorders. To further illustrate its clinical potential, the method was applied to differentiating in vivo OCT esophagus images from an eosinophilic esophagitis (EOE) model and its control group, and the results clearly demonstrated quantitative changes in the top esophageal layers’ thickness in the EOE model.",
author = "Dawei Li and Jimin Wu and Yufan He and Xinwen Yao and Wu Yuan and Defu Chen and Park, {Hyeon Cheol} and Shaoyong Yu and Prince, {Jerry Ladd} and Xingde Li",
year = "2019",
month = "3",
day = "1",
doi = "10.1364/BOE.10.001126",
language = "English (US)",
volume = "10",
pages = "1126--1135",
journal = "Biomedical Optics Express",
issn = "2156-7085",
publisher = "The Optical Society",
number = "3",

}

TY - JOUR

T1 - Parallel deep neural networks for endoscopic oct image segmentation

AU - Li, Dawei

AU - Wu, Jimin

AU - He, Yufan

AU - Yao, Xinwen

AU - Yuan, Wu

AU - Chen, Defu

AU - Park, Hyeon Cheol

AU - Yu, Shaoyong

AU - Prince, Jerry Ladd

AU - Li, Xingde

PY - 2019/3/1

Y1 - 2019/3/1

N2 - We report parallel-trained deep neural networks for automated endoscopic OCT image segmentation feasible even with a limited training data set. These U-Net-based deep neural networks were trained using a modified dice loss function and manual segmentations of ultrahigh-resolution cross-sectional images collected by an 800 nm OCT endoscopic system. The method was tested on in vivo guinea pig esophagus images. Results showed its robust layer segmentation capability with a boundary error of 1.4 µm insensitive to lay topology disorders. To further illustrate its clinical potential, the method was applied to differentiating in vivo OCT esophagus images from an eosinophilic esophagitis (EOE) model and its control group, and the results clearly demonstrated quantitative changes in the top esophageal layers’ thickness in the EOE model.

AB - We report parallel-trained deep neural networks for automated endoscopic OCT image segmentation feasible even with a limited training data set. These U-Net-based deep neural networks were trained using a modified dice loss function and manual segmentations of ultrahigh-resolution cross-sectional images collected by an 800 nm OCT endoscopic system. The method was tested on in vivo guinea pig esophagus images. Results showed its robust layer segmentation capability with a boundary error of 1.4 µm insensitive to lay topology disorders. To further illustrate its clinical potential, the method was applied to differentiating in vivo OCT esophagus images from an eosinophilic esophagitis (EOE) model and its control group, and the results clearly demonstrated quantitative changes in the top esophageal layers’ thickness in the EOE model.

UR - http://www.scopus.com/inward/record.url?scp=85063947655&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85063947655&partnerID=8YFLogxK

U2 - 10.1364/BOE.10.001126

DO - 10.1364/BOE.10.001126

M3 - Article

C2 - 30891334

AN - SCOPUS:85063947655

VL - 10

SP - 1126

EP - 1135

JO - Biomedical Optics Express

JF - Biomedical Optics Express

SN - 2156-7085

IS - 3

M1 - #349139

ER -