About this item:

111 Views | 56 Downloads

Author Notes:

Michael A. Marchetti, MD, Dermatology Service, Department of Medicine, Memorial Sloan Kettering Cancer Center, 16 East 60th Street, New York, NY 10022, U.S.A. Telephone: 646-888-6016. Fax: 646-227-7274. Email: marchetm@mskcc.org

This study reports the efforts of the International Skin Imaging Collaboration (ISIC). We thank the leadership of the ISIC collaboration: H. Peter Soyer, MD, Dermatology Research Centre, The University of Queensland, Brisbane, Australia (Technique Working Group Co-Leader); Clara Curiel-Lewandrowski, MD, University of Arizona Cancer Center, Tucson, AZ, USA (Technique Working Group Co-Leader); Harald Kittler, MD, Department of Dermatology, Medical University of Vienna, Austria (Terminology Working Group Leader); Liam Caffery, PhD, The University of Queensland, Brisbane, Australia (Metadata Working Group Leader); Josep Malvehy, MD; Hospital Clinic of Barcelona, Spain (Technology Working Group Leader); Rainer Hofmann Wellenhof, MD, Medical University of Graz, Austria (Archive Group Leader).

The authors also thank the organizing committee of the 2017 International Symposium on Biomedical Imaging (ISBI), the chairs of the 2017 ISBI Grand Challenges: Bram van Ginneken, Radboud University Medical Center, NL; Adriënne Mendrik, Utrecht University, NL; Stephen Aylward, Kitware Inc., USA, the participants of the 2017 ISBI Challenge “Skin Lesion Analysis towards Melanoma Detection, and the participants of the reader study: Cristina Carrera, MD, PhD; Jennifer L. DeFazio, MD; Natalia Jaimes, MD; Ashfaq A. Marghoob, MD; Elizabeth Quigley, MD; Alon Scope, MD; Oriol Yelamos, MD; Allan C. Halpern, MD; Caren Waintraub, MD; Meryl Rosen, MD; Sarah Jawed, MD; Priyanka Gumaste, MD; Miriam R. Lieberman, MD; Silvia Mancebo, MD; Christine Totri, MD; Corey Georgesen, MD; Freya Van Driessche, MD; Maira Fonseca, MD.

Dr. Codella is an employee of IBM and an IBM stockholder. Dr. Halpern is a consultant for Canfield Scientific Inc, Caliber I.D., and SciBase. Drs. Marchetti, Liopyris, Kalloo, Gutman, Helba, and Dusza have no financial disclosures

Subjects:

Research Funding:

This research was funded in part through the NIH/NCI Cancer Center Support Grant P30 CA008748.

Keywords:

  • Science & Technology
  • Life Sciences & Biomedicine
  • Dermatology
  • automated melanoma diagnosis
  • computer algorithm
  • computer vision
  • deep learning
  • dermatologist
  • International Skin Imaging Collaboration
  • International Symposium on Biomedical Imaging
  • machine learning
  • melanoma
  • reader study
  • skin cancer

Computer algorithms show potential for improving dermatologists' accuracy to diagnose cutaneous melanoma: Results of the International Skin Imaging Collaboration 2017

Tools:

Journal Title:

JOURNAL OF THE AMERICAN ACADEMY OF DERMATOLOGY

Volume:

Volume 82, Number 3

Publisher:

, Pages 622-627

Type of Work:

Article | Post-print: After Peer Review

Abstract:

Background: Computer vision has promise in image-based cutaneous melanoma diagnosis but clinical utility is uncertain. Objective: To determine if computer algorithms from an international melanoma detection challenge can improve dermatologists' accuracy in diagnosing melanoma. Methods: In this cross-sectional study, we used 150 dermoscopy images (50 melanomas, 50 nevi, 50 seborrheic keratoses) from the test dataset of a melanoma detection challenge, along with algorithm results from 23 teams. Eight dermatologists and 9 dermatology residents classified dermoscopic lesion images in an online reader study and provided their confidence level. Results: The top-ranked computer algorithm had an area under the receiver operating characteristic curve of 0.87, which was higher than that of the dermatologists (0.74) and residents (0.66) (P <. 001 for all comparisons). At the dermatologists' overall sensitivity in classification of 76.0%, the algorithm had a superior specificity (85.0% vs. 72.6%, P = .001). Imputation of computer algorithm classifications into dermatologist evaluations with low confidence ratings (26.6% of evaluations) increased dermatologist sensitivity from 76.0% to 80.8% and specificity from 72.6% to 72.8%. Limitations: Artificial study setting lacking the full spectrum of skin lesions as well as clinical metadata. Conclusion: Accumulating evidence suggests that deep neural networks can classify skin images of melanoma and its benign mimickers with high accuracy and potentially improve human performance.

Copyright information:

This is an Open Access work distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License (https://creativecommons.org/licenses/by-nc-nd/4.0/rdf).
Export to EndNote