Computer-aided diagnosis improves characterization of Barrett's neoplasia by general endoscopists (with video)

BONS-AI consortium

doi:10.1016/j.gie.2024.04.013

Computer-aided diagnosis improves characterization of Barrett's neoplasia by general endoscopists (with video)

BONS-AI consortium

Medicine, Gastroenterology Division

Research output: Contribution to journal › Article › peer-review

3 Scopus citations

Abstract

Background and Aims: Characterization of visible abnormalities in patients with Barrett's esophagus (BE) can be challenging, especially for inexperienced endoscopists. This results in suboptimal diagnostic accuracy and poor interobserver agreement. Computer-aided diagnosis (CADx) systems may assist endoscopists. We aimed to develop, validate, and benchmark a CADx system for BE neoplasia. Methods: The CADx system received pretraining with ImageNet and then consecutive domain-specific pretraining with GastroNet, which includes 5 million endoscopic images. It was subsequently trained and internally validated using 1758 narrow-band imaging (NBI) images of early BE neoplasia (352 patients) and 1838 NBI images of nondysplastic BE (173 patients) from 8 international centers. CADx was tested prospectively on corresponding image and video test sets with 30 cases (20 patients) of BE neoplasia and 60 cases (31 patients) of nondysplastic BE. The test set was benchmarked by 44 general endoscopists in 2 phases (phase 1, no CADx assistance; phase 2, with CADx assistance). Ten international BE experts provided additional benchmark performance. Results: Stand-alone sensitivity and specificity of the CADx system were 100% and 98% for images and 93% and 96% for videos, respectively. CADx outperformed general endoscopists without CADx assistance in terms of sensitivity (P = .04). Sensitivity and specificity of general endoscopists increased from 84% to 96% and 90% to 98% with CAD assistance (P < .001). CADx assistance increased endoscopists’ confidence in characterization (P < .001). CADx performance was similar to that of the BE experts. Conclusions: CADx assistance significantly increased characterization performance of BE neoplasia by general endoscopists to the level of expert endoscopists. The use of this CADx system may thereby improve daily Barrett surveillance.

Original language	English (US)
Pages (from-to)	616-625.e8
Journal	Gastrointestinal endoscopy
Volume	100
Issue number	4
DOIs	https://doi.org/10.1016/j.gie.2024.04.013
State	Published - Oct 2024

Funding

The authors thank the members and collaborators of the BONS-AI consortium, as listed below in alphabetical order: Alaa Alkhalaf, Isala Hospital, Zwolle, the Netherlands; Lorenza Alvarez Herrero, St. Antonius Hospital, Utrecht, the Netherlands; Francisco Baldaque-Silva, Karolinska University Hospital, Stockholm, Sweden; Maximilien Barret, Cochin Hospital, Paris, France; Jacques J. Bergman, Amsterdam UMC, Amsterdam, the Netherlands; Torsten Beyna, Evangelisches Krankenhaus D\u00FCsseldorf, D\u00FCsseldorf, Germany; Raf Bisschops, University Hospitals Leuven, Leuven, Belgium; Tim G. Boers, TU Eindhoven, Eindhoven, the Netherlands; Wouter Curvers, Catharina Hospital, Eindhoven, the Netherlands; Pierre H. Deprez, University Hospital Saint-Luc, Brussels, Belgium; Lucas C. Duits, Amsterdam UMC, Amsterdam, the Netherlands; Peter Elbe, Karolinska University Hospital & Karolinska Institute, Stockholm, Sweden; Jose Miguel Esteban, Cl\u00EDnico San Carlos, Madrid, Spain; Gary W. Falk, Hospital of the University of Pennsylvania, Philadelphia, Pennsylvania, USA; Kiki N. Fockens, Amsterdam UMC, Amsterdam, the Netherlands; Gregory G. Ginsberg, University of Pennsylvania Perelman School of Medicine, Philadelphia, Pennsylvania, USA; Albert J de Groof, Amsterdam UMC, Amsterdam, the Netherlands; Rehan Haidry, Cleveland Clinic London and University College Hospitals NHS, London, United Kingdom; Martin H. Houben, HagaZiekenhuis, The Hague, the Netherlands; Anthony Infantolino, Thomas Jefferson University Hospital, Philadelphia, Pennsylvania, USA; Prasad G. Iyer, Mayo Clinic, Rochester, Minnesota, USA; Martijn Jong, Amsterdam UMC, Amsterdam, the Netherlands; Pieter-Jan de Jonge, Erasmus MC, Rotterdam, the Netherlands; Jelmer B. Jukema, Amsterdam UMC, Amsterdam the Netherlands; Arjun Koch, Erasmus MC, Rotterdam, the Netherlands; Srinadh Komanduri, Feinberg School of Medicine, Northwestern University, Chicago, Illinois, USA; Vani Konda, Baylor University Medical Center, Dallas, Texas, USA; Koen C. Kusters, TU Eindhoven, Eindhoven, the Netherlands; Philippe Leclerq, University Hospitals Leuven, Leuven, Belgium; Cadman L. Leggett, Mayo Clinic, Rochester, Minnesota, USA; Charles J. Lightdale, Columbia University Medical Center, New York, New York, USA; Rosalie C. Mallant-Hent, Flevoziekenhuis, Almere, the Netherlands; Guiomar Moral Villarejo, Nottingham University Hospitals NHS Trust, Nottingham, United Kingdom. V. Raman Muthusamy, UCLA Health, Los Angeles, California, USA; Jacobo Ortiz Fern\u00E1ndez-Sordo, Nottingham University Hospitals NHS Trust, Nottingham, United Kingdom; Oliver Pech, St. John of God Hospital, Regensburg, Germany; Ian Penman, Royal Infirmary Edinburgh, Edinburgh, United Kingdom; Roos E. Pouw, Amsterdam UMC, Amsterdam, the Netherlands; Joost A. van der Putten, TU Eindhoven, Eindhoven, the Netherlands; Krish Ragunath, Royal Perth Hospital, Perth, Australia; Pieter Scholten, Onze Lieve Vrouwe Gasthuis, Amsterdam, the Netherlands; Stefan Seewald, Klinik Hirslanden, Zurich, Switzerland; Amritha Sethi, Columbia University Medical Center, New York, New York, USA; Michael S. Smith, Mount Sinai West & Mount Sinai Morningside Hospitals, New York, New York, USA; Fons van der Sommen, TU Eindhoven, Eindhoven, the Netherlands; Arvind Trindade, Zucker School of Medicine at Hofstra/Northwell, Long Island Jewish Medical Center, New Hyde Park, New York, New York, USA; Sachin Wani, University of Colorado Anschutz Medical Center, Aurora, Colorado, USA; Irving Waxman, Rush University Medical Center, Chicago, Illinois, USA; Jessie Westerhof, UMC Groningen, Groningen, the Netherlands; Bas L. Weusten, UMC Utrecht & St. Antonius Hospital, Utrecht, the Netherlands; Peter H. de With, TU Eindhoven, Eindhoven, the Netherlands; and Herbert C. Wolfsen, Mayo Clinic, Jacksonville, Florida, USA.

ASJC Scopus subject areas

Radiology Nuclear Medicine and imaging
Gastroenterology

Access to Document

10.1016/j.gie.2024.04.013

Cite this

@article{4177df03cad54eb59b4066aa8d5b0906,

title = "Computer-aided diagnosis improves characterization of Barrett's neoplasia by general endoscopists (with video)",

abstract = "Background and Aims: Characterization of visible abnormalities in patients with Barrett's esophagus (BE) can be challenging, especially for inexperienced endoscopists. This results in suboptimal diagnostic accuracy and poor interobserver agreement. Computer-aided diagnosis (CADx) systems may assist endoscopists. We aimed to develop, validate, and benchmark a CADx system for BE neoplasia. Methods: The CADx system received pretraining with ImageNet and then consecutive domain-specific pretraining with GastroNet, which includes 5 million endoscopic images. It was subsequently trained and internally validated using 1758 narrow-band imaging (NBI) images of early BE neoplasia (352 patients) and 1838 NBI images of nondysplastic BE (173 patients) from 8 international centers. CADx was tested prospectively on corresponding image and video test sets with 30 cases (20 patients) of BE neoplasia and 60 cases (31 patients) of nondysplastic BE. The test set was benchmarked by 44 general endoscopists in 2 phases (phase 1, no CADx assistance; phase 2, with CADx assistance). Ten international BE experts provided additional benchmark performance. Results: Stand-alone sensitivity and specificity of the CADx system were 100% and 98% for images and 93% and 96% for videos, respectively. CADx outperformed general endoscopists without CADx assistance in terms of sensitivity (P = .04). Sensitivity and specificity of general endoscopists increased from 84% to 96% and 90% to 98% with CAD assistance (P < .001). CADx assistance increased endoscopists{\textquoteright} confidence in characterization (P < .001). CADx performance was similar to that of the BE experts. Conclusions: CADx assistance significantly increased characterization performance of BE neoplasia by general endoscopists to the level of expert endoscopists. The use of this CADx system may thereby improve daily Barrett surveillance.",

author = "{BONS-AI consortium} and Jukema, {Jelmer B.} and Kusters, {Carolus H.J.} and Jong, {Martijn R.} and Fockens, {Kiki N.} and Tim Boers and {van der Putten}, {Joost A.} and Pouw, {Roos E.} and Duits, {Lucas C.} and Weusten, {Bas L.A.M.} and Herrero, {Lorenza Alvarez} and Houben, {Martin H.M.G.} and Nagengast, {Wouter B.} and Jessie Westerhof and Alaa Alkhalaf and Rosalie Mallant-Hent and Pieter Scholten and Krish Ragunath and Stefan Seewald and Peter Elbe and Silva, {Francisco Baldaque} and Maximilien Barret and {Ortiz Fern{\'a}ndez-Sordo}, Jacobo and {Moral Villarejo}, Guiomar and Oliver Pech and Torsten Beyna and Montazeri, {Nahid S.M.} and {der Sommen}, {Fons van} and {de With}, {Peter H.} and {de Groof}, {A. Jeroen} and Bergman, {Jacques J.} and Francisco Baldaque-Silva and Raf Bisschops and Boers, {Tim G.} and Wouter Curvers and Deprez, {Pierre H.} and Esteban, {Jose Miguel} and Falk, {Gary W.} and Ginsberg, {Gregory G.} and {de Groof}, {Albert J.} and Rehan Haidry and Houben, {Martin H.} and Anthony Infantolino and Iyer, {Prasad G.} and Martijn Jong and Jonge, {Pieter Jan de} and Arjun Koch and Srinadh Komanduri and Vani Konda and Kusters, {Koen C.} and Philippe Leclerq",

note = "Publisher Copyright: {\textcopyright} 2024 American Society for Gastrointestinal Endoscopy",

year = "2024",

month = oct,

doi = "10.1016/j.gie.2024.04.013",

language = "English (US)",

volume = "100",

pages = "616--625.e8",

journal = "Gastrointestinal endoscopy",

issn = "0016-5107",

publisher = "Elsevier Inc.",

number = "4",

}

TY - JOUR

T1 - Computer-aided diagnosis improves characterization of Barrett's neoplasia by general endoscopists (with video)

AU - BONS-AI consortium

AU - Jukema, Jelmer B.

AU - Kusters, Carolus H.J.

AU - Jong, Martijn R.

AU - Fockens, Kiki N.

AU - Boers, Tim

AU - van der Putten, Joost A.

AU - Pouw, Roos E.

AU - Duits, Lucas C.

AU - Weusten, Bas L.A.M.

AU - Herrero, Lorenza Alvarez

AU - Houben, Martin H.M.G.

AU - Nagengast, Wouter B.

AU - Westerhof, Jessie

AU - Alkhalaf, Alaa

AU - Mallant-Hent, Rosalie

AU - Scholten, Pieter

AU - Ragunath, Krish

AU - Seewald, Stefan

AU - Elbe, Peter

AU - Silva, Francisco Baldaque

AU - Barret, Maximilien

AU - Ortiz Fernández-Sordo, Jacobo

AU - Moral Villarejo, Guiomar

AU - Pech, Oliver

AU - Beyna, Torsten

AU - Montazeri, Nahid S.M.

AU - der Sommen, Fons van

AU - de With, Peter H.

AU - de Groof, A. Jeroen

AU - Bergman, Jacques J.

AU - Baldaque-Silva, Francisco

AU - Bisschops, Raf

AU - Boers, Tim G.

AU - Curvers, Wouter

AU - Deprez, Pierre H.

AU - Esteban, Jose Miguel

AU - Falk, Gary W.

AU - Ginsberg, Gregory G.

AU - de Groof, Albert J.

AU - Haidry, Rehan

AU - Houben, Martin H.

AU - Infantolino, Anthony

AU - Iyer, Prasad G.

AU - Jong, Martijn

AU - Jonge, Pieter Jan de

AU - Koch, Arjun

AU - Komanduri, Srinadh

AU - Konda, Vani

AU - Kusters, Koen C.

AU - Leclerq, Philippe

PY - 2024/10

Y1 - 2024/10

N2 - Background and Aims: Characterization of visible abnormalities in patients with Barrett's esophagus (BE) can be challenging, especially for inexperienced endoscopists. This results in suboptimal diagnostic accuracy and poor interobserver agreement. Computer-aided diagnosis (CADx) systems may assist endoscopists. We aimed to develop, validate, and benchmark a CADx system for BE neoplasia. Methods: The CADx system received pretraining with ImageNet and then consecutive domain-specific pretraining with GastroNet, which includes 5 million endoscopic images. It was subsequently trained and internally validated using 1758 narrow-band imaging (NBI) images of early BE neoplasia (352 patients) and 1838 NBI images of nondysplastic BE (173 patients) from 8 international centers. CADx was tested prospectively on corresponding image and video test sets with 30 cases (20 patients) of BE neoplasia and 60 cases (31 patients) of nondysplastic BE. The test set was benchmarked by 44 general endoscopists in 2 phases (phase 1, no CADx assistance; phase 2, with CADx assistance). Ten international BE experts provided additional benchmark performance. Results: Stand-alone sensitivity and specificity of the CADx system were 100% and 98% for images and 93% and 96% for videos, respectively. CADx outperformed general endoscopists without CADx assistance in terms of sensitivity (P = .04). Sensitivity and specificity of general endoscopists increased from 84% to 96% and 90% to 98% with CAD assistance (P < .001). CADx assistance increased endoscopists’ confidence in characterization (P < .001). CADx performance was similar to that of the BE experts. Conclusions: CADx assistance significantly increased characterization performance of BE neoplasia by general endoscopists to the level of expert endoscopists. The use of this CADx system may thereby improve daily Barrett surveillance.

AB - Background and Aims: Characterization of visible abnormalities in patients with Barrett's esophagus (BE) can be challenging, especially for inexperienced endoscopists. This results in suboptimal diagnostic accuracy and poor interobserver agreement. Computer-aided diagnosis (CADx) systems may assist endoscopists. We aimed to develop, validate, and benchmark a CADx system for BE neoplasia. Methods: The CADx system received pretraining with ImageNet and then consecutive domain-specific pretraining with GastroNet, which includes 5 million endoscopic images. It was subsequently trained and internally validated using 1758 narrow-band imaging (NBI) images of early BE neoplasia (352 patients) and 1838 NBI images of nondysplastic BE (173 patients) from 8 international centers. CADx was tested prospectively on corresponding image and video test sets with 30 cases (20 patients) of BE neoplasia and 60 cases (31 patients) of nondysplastic BE. The test set was benchmarked by 44 general endoscopists in 2 phases (phase 1, no CADx assistance; phase 2, with CADx assistance). Ten international BE experts provided additional benchmark performance. Results: Stand-alone sensitivity and specificity of the CADx system were 100% and 98% for images and 93% and 96% for videos, respectively. CADx outperformed general endoscopists without CADx assistance in terms of sensitivity (P = .04). Sensitivity and specificity of general endoscopists increased from 84% to 96% and 90% to 98% with CAD assistance (P < .001). CADx assistance increased endoscopists’ confidence in characterization (P < .001). CADx performance was similar to that of the BE experts. Conclusions: CADx assistance significantly increased characterization performance of BE neoplasia by general endoscopists to the level of expert endoscopists. The use of this CADx system may thereby improve daily Barrett surveillance.

UR - http://www.scopus.com/inward/record.url?scp=85201584235&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85201584235&partnerID=8YFLogxK

U2 - 10.1016/j.gie.2024.04.013

DO - 10.1016/j.gie.2024.04.013

M3 - Article

C2 - 38636819

AN - SCOPUS:85201584235

SN - 0016-5107

VL - 100

SP - 616-625.e8

JO - Gastrointestinal endoscopy

JF - Gastrointestinal endoscopy

IS - 4

ER -

Computer-aided diagnosis improves characterization of Barrett's neoplasia by general endoscopists (with video)

Abstract

Funding

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this