Tissue segmentation in volumetric laser endomicroscopy data using FusionNet and a domain-specific loss function

Joost van der Putten, Fons van der Sommen, Maarten Struyvenberg, Jeroen de Groof, Wouter Curvers, Erik Schoon, Jacques J.G.H.M. Bergman, Peter H.N. de With

Research output: Contribution to journalConference articleAcademicpeer-review

Abstract

Volumetric Laser Endomicroscopy (VLE) is a promising balloon-based imaging technique for detecting early neoplasia in Barretts Esophagus. An essential pre-processing step for further processing is tissue segmentation. At present, tissue is segmented manually but is not scalable for the entire VLE scan consisting of 1,200 frames of 4,096 × 2,048 pixels. This paper explores the possibility of automatically segmenting relevant tissue for VLE scans using FusionNet and a domain-specific loss function. The contribution of this work is threefold. First, we propose a tissue segmentation algorithm for VLE scans. Second, we introduce a weighted ground truth that exploits the signal-to-noise ratio characteristics of the VLE data. Third, we compare our algorithm segmentation against two additional VLE experts. The results show that our algorithm annotations are indistinguishable from the expert annotations and therefore the algorithm can be used as a preprocessing step for further classification.

Fingerprint

Loss Function
Segmentation
Tissue
Laser
Lasers
lasers
annotations
preprocessing
Preprocessing
Annotation
esophagus
Balloon
ground truth
Balloons
balloons
Threefolds
Processing
imaging techniques
Signal to noise ratio
signal to noise ratios

Cite this

van der Putten, Joost ; van der Sommen, Fons ; Struyvenberg, Maarten ; de Groof, Jeroen ; Curvers, Wouter ; Schoon, Erik ; J.G.H.M. Bergman, Jacques ; de With, Peter H.N./ Tissue segmentation in volumetric laser endomicroscopy data using FusionNet and a domain-specific loss function. In: Proceedings of SPIE. 2019 ; Vol. 10949.
@article{5b12b73a7e964ce4b95ddbae3783e064,
title = "Tissue segmentation in volumetric laser endomicroscopy data using FusionNet and a domain-specific loss function",
abstract = "Volumetric Laser Endomicroscopy (VLE) is a promising balloon-based imaging technique for detecting early neoplasia in Barretts Esophagus. An essential pre-processing step for further processing is tissue segmentation. At present, tissue is segmented manually but is not scalable for the entire VLE scan consisting of 1,200 frames of 4,096 × 2,048 pixels. This paper explores the possibility of automatically segmenting relevant tissue for VLE scans using FusionNet and a domain-specific loss function. The contribution of this work is threefold. First, we propose a tissue segmentation algorithm for VLE scans. Second, we introduce a weighted ground truth that exploits the signal-to-noise ratio characteristics of the VLE data. Third, we compare our algorithm segmentation against two additional VLE experts. The results show that our algorithm annotations are indistinguishable from the expert annotations and therefore the algorithm can be used as a preprocessing step for further classification.",
author = "{van der Putten}, Joost and {van der Sommen}, Fons and Maarten Struyvenberg and {de Groof}, Jeroen and Wouter Curvers and Erik Schoon and {J.G.H.M. Bergman}, Jacques and {de With}, {Peter H.N.}",
year = "2019",
doi = "10.1117/12.2512192",
language = "English",
volume = "10949",
journal = "Proceedings of SPIE",
issn = "0277-786X",
publisher = "SPIE",

}

Tissue segmentation in volumetric laser endomicroscopy data using FusionNet and a domain-specific loss function. / van der Putten, Joost; van der Sommen, Fons; Struyvenberg, Maarten; de Groof, Jeroen; Curvers, Wouter; Schoon, Erik; J.G.H.M. Bergman, Jacques; de With, Peter H.N.

In: Proceedings of SPIE, Vol. 10949, 109492J, 2019.

Research output: Contribution to journalConference articleAcademicpeer-review

TY - JOUR

T1 - Tissue segmentation in volumetric laser endomicroscopy data using FusionNet and a domain-specific loss function

AU - van der Putten,Joost

AU - van der Sommen,Fons

AU - Struyvenberg,Maarten

AU - de Groof,Jeroen

AU - Curvers,Wouter

AU - Schoon,Erik

AU - J.G.H.M. Bergman,Jacques

AU - de With,Peter H.N.

PY - 2019

Y1 - 2019

N2 - Volumetric Laser Endomicroscopy (VLE) is a promising balloon-based imaging technique for detecting early neoplasia in Barretts Esophagus. An essential pre-processing step for further processing is tissue segmentation. At present, tissue is segmented manually but is not scalable for the entire VLE scan consisting of 1,200 frames of 4,096 × 2,048 pixels. This paper explores the possibility of automatically segmenting relevant tissue for VLE scans using FusionNet and a domain-specific loss function. The contribution of this work is threefold. First, we propose a tissue segmentation algorithm for VLE scans. Second, we introduce a weighted ground truth that exploits the signal-to-noise ratio characteristics of the VLE data. Third, we compare our algorithm segmentation against two additional VLE experts. The results show that our algorithm annotations are indistinguishable from the expert annotations and therefore the algorithm can be used as a preprocessing step for further classification.

AB - Volumetric Laser Endomicroscopy (VLE) is a promising balloon-based imaging technique for detecting early neoplasia in Barretts Esophagus. An essential pre-processing step for further processing is tissue segmentation. At present, tissue is segmented manually but is not scalable for the entire VLE scan consisting of 1,200 frames of 4,096 × 2,048 pixels. This paper explores the possibility of automatically segmenting relevant tissue for VLE scans using FusionNet and a domain-specific loss function. The contribution of this work is threefold. First, we propose a tissue segmentation algorithm for VLE scans. Second, we introduce a weighted ground truth that exploits the signal-to-noise ratio characteristics of the VLE data. Third, we compare our algorithm segmentation against two additional VLE experts. The results show that our algorithm annotations are indistinguishable from the expert annotations and therefore the algorithm can be used as a preprocessing step for further classification.

U2 - 10.1117/12.2512192

DO - 10.1117/12.2512192

M3 - Conference article

VL - 10949

JO - Proceedings of SPIE

T2 - Proceedings of SPIE

JF - Proceedings of SPIE

SN - 0277-786X

M1 - 109492J

ER -