Sweet-spot training for early esophageal cancer detection

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

3 Citations (Scopus)
3 Downloads (Pure)

Abstract

Over the past decade, the imaging tools for endoscopists have improved drastically. This has enabled physicians to visually inspect the intestinal tissue for early signs of malignant lesions. Furthermore, it has paved the way for image analysis algorithms, to support the gastroenterologist in finding these early signs of developing cancer. Recent studies show the feasibility of such systems, where the problem is typically approached as a segmentation task and a binary ground truth is employed. In this study, we show that the detection of early cancerous tissue in the gastrointestinal tract cannot be approached as a binary segmentation problem and it is crucial and clinically relevant to involve multiple experts for annotating early lesions. By employing the so-called sweet spot for training purposes, a much better detection performance can be achieved. Furthermore, a multi-expert-based ground truth, i.e. a golden standard, enables an improved validation of the resulting delineations. For this purpose, we propose two novel metrics that can handle multiple ground-truth annotations. Our experiments involving these metrics and based on the golden standard show that the performance of a detection algorithm of early neoplastic lesions in Barrett's esophagus can be increased significantly, demonstrating a 10 percent point increase in the resulting F1 detection score.
Original languageEnglish
Title of host publicationMedical Imaging 2016 : Computer-Aided Diagnosis, February 27th - March 3rd 2016, San Diego, California, USA
EditorsG.D. Tourassi, S.G. Armato
PublisherSPIE
Pages1-7
ISBN (Print)9781510600201
DOIs
Publication statusPublished - 2016
Event2016 Medical Imaging - San Diego, United States
Duration: 27 Feb 20163 Mar 2016

Publication series

NameProceedings of SPIE
Volume9785
ISSN (Print)0277-786X

Conference

Conference2016 Medical Imaging
CountryUnited States
CitySan Diego
Period27/02/163/03/16
Other"Computer-Aided Diagnosis"

Fingerprint

Tissue
Image analysis
Imaging techniques
Experiments

Keywords

  • Computer-Aided Diagnosis
  • Esophageal cancer
  • Shape similarity
  • Multi-expert validation

Cite this

van der Sommen, F., Zinger, S., Schoon, E. J., & de With, P. H. N. (2016). Sweet-spot training for early esophageal cancer detection. In G. D. Tourassi, & S. G. Armato (Eds.), Medical Imaging 2016 : Computer-Aided Diagnosis, February 27th - March 3rd 2016, San Diego, California, USA (pp. 1-7). [IB] (Proceedings of SPIE; Vol. 9785). SPIE. https://doi.org/10.1117/12.2208114
van der Sommen, F. ; Zinger, S. ; Schoon, E.J. ; de With, P.H.N. / Sweet-spot training for early esophageal cancer detection. Medical Imaging 2016 : Computer-Aided Diagnosis, February 27th - March 3rd 2016, San Diego, California, USA. editor / G.D. Tourassi ; S.G. Armato. SPIE, 2016. pp. 1-7 (Proceedings of SPIE).
@inproceedings{4356ac6e3e424473adb80f164c8b1562,
title = "Sweet-spot training for early esophageal cancer detection",
abstract = "Over the past decade, the imaging tools for endoscopists have improved drastically. This has enabled physicians to visually inspect the intestinal tissue for early signs of malignant lesions. Furthermore, it has paved the way for image analysis algorithms, to support the gastroenterologist in finding these early signs of developing cancer. Recent studies show the feasibility of such systems, where the problem is typically approached as a segmentation task and a binary ground truth is employed. In this study, we show that the detection of early cancerous tissue in the gastrointestinal tract cannot be approached as a binary segmentation problem and it is crucial and clinically relevant to involve multiple experts for annotating early lesions. By employing the so-called sweet spot for training purposes, a much better detection performance can be achieved. Furthermore, a multi-expert-based ground truth, i.e. a golden standard, enables an improved validation of the resulting delineations. For this purpose, we propose two novel metrics that can handle multiple ground-truth annotations. Our experiments involving these metrics and based on the golden standard show that the performance of a detection algorithm of early neoplastic lesions in Barrett's esophagus can be increased significantly, demonstrating a 10 percent point increase in the resulting F1 detection score.",
keywords = "Computer-Aided Diagnosis, Esophageal cancer, Shape similarity, Multi-expert validation",
author = "{van der Sommen}, F. and S. Zinger and E.J. Schoon and {de With}, P.H.N.",
year = "2016",
doi = "10.1117/12.2208114",
language = "English",
isbn = "9781510600201",
series = "Proceedings of SPIE",
publisher = "SPIE",
pages = "1--7",
editor = "G.D. Tourassi and S.G. Armato",
booktitle = "Medical Imaging 2016 : Computer-Aided Diagnosis, February 27th - March 3rd 2016, San Diego, California, USA",
address = "United States",

}

van der Sommen, F, Zinger, S, Schoon, EJ & de With, PHN 2016, Sweet-spot training for early esophageal cancer detection. in GD Tourassi & SG Armato (eds), Medical Imaging 2016 : Computer-Aided Diagnosis, February 27th - March 3rd 2016, San Diego, California, USA., IB, Proceedings of SPIE, vol. 9785, SPIE, pp. 1-7, 2016 Medical Imaging, San Diego, United States, 27/02/16. https://doi.org/10.1117/12.2208114

Sweet-spot training for early esophageal cancer detection. / van der Sommen, F.; Zinger, S.; Schoon, E.J.; de With, P.H.N.

Medical Imaging 2016 : Computer-Aided Diagnosis, February 27th - March 3rd 2016, San Diego, California, USA. ed. / G.D. Tourassi; S.G. Armato. SPIE, 2016. p. 1-7 IB (Proceedings of SPIE; Vol. 9785).

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

TY - GEN

T1 - Sweet-spot training for early esophageal cancer detection

AU - van der Sommen, F.

AU - Zinger, S.

AU - Schoon, E.J.

AU - de With, P.H.N.

PY - 2016

Y1 - 2016

N2 - Over the past decade, the imaging tools for endoscopists have improved drastically. This has enabled physicians to visually inspect the intestinal tissue for early signs of malignant lesions. Furthermore, it has paved the way for image analysis algorithms, to support the gastroenterologist in finding these early signs of developing cancer. Recent studies show the feasibility of such systems, where the problem is typically approached as a segmentation task and a binary ground truth is employed. In this study, we show that the detection of early cancerous tissue in the gastrointestinal tract cannot be approached as a binary segmentation problem and it is crucial and clinically relevant to involve multiple experts for annotating early lesions. By employing the so-called sweet spot for training purposes, a much better detection performance can be achieved. Furthermore, a multi-expert-based ground truth, i.e. a golden standard, enables an improved validation of the resulting delineations. For this purpose, we propose two novel metrics that can handle multiple ground-truth annotations. Our experiments involving these metrics and based on the golden standard show that the performance of a detection algorithm of early neoplastic lesions in Barrett's esophagus can be increased significantly, demonstrating a 10 percent point increase in the resulting F1 detection score.

AB - Over the past decade, the imaging tools for endoscopists have improved drastically. This has enabled physicians to visually inspect the intestinal tissue for early signs of malignant lesions. Furthermore, it has paved the way for image analysis algorithms, to support the gastroenterologist in finding these early signs of developing cancer. Recent studies show the feasibility of such systems, where the problem is typically approached as a segmentation task and a binary ground truth is employed. In this study, we show that the detection of early cancerous tissue in the gastrointestinal tract cannot be approached as a binary segmentation problem and it is crucial and clinically relevant to involve multiple experts for annotating early lesions. By employing the so-called sweet spot for training purposes, a much better detection performance can be achieved. Furthermore, a multi-expert-based ground truth, i.e. a golden standard, enables an improved validation of the resulting delineations. For this purpose, we propose two novel metrics that can handle multiple ground-truth annotations. Our experiments involving these metrics and based on the golden standard show that the performance of a detection algorithm of early neoplastic lesions in Barrett's esophagus can be increased significantly, demonstrating a 10 percent point increase in the resulting F1 detection score.

KW - Computer-Aided Diagnosis

KW - Esophageal cancer

KW - Shape similarity

KW - Multi-expert validation

U2 - 10.1117/12.2208114

DO - 10.1117/12.2208114

M3 - Conference contribution

SN - 9781510600201

T3 - Proceedings of SPIE

SP - 1

EP - 7

BT - Medical Imaging 2016 : Computer-Aided Diagnosis, February 27th - March 3rd 2016, San Diego, California, USA

A2 - Tourassi, G.D.

A2 - Armato, S.G.

PB - SPIE

ER -

van der Sommen F, Zinger S, Schoon EJ, de With PHN. Sweet-spot training for early esophageal cancer detection. In Tourassi GD, Armato SG, editors, Medical Imaging 2016 : Computer-Aided Diagnosis, February 27th - March 3rd 2016, San Diego, California, USA. SPIE. 2016. p. 1-7. IB. (Proceedings of SPIE). https://doi.org/10.1117/12.2208114