Systematic outperformance of 112 dermatologists in multiclass skin cancer image classification by convolutional neural networks

Maron, Roman C. ; Weichenthal, Michael ; Utikal, Jochen S. ; Hekler, Achim ; Berking, Carola ; Hauschild, Axel ; Enk, Alexander H. ; Haferkamp, Sebastian ; Klode, Joachim ; Schadendorf, Dirk

; Jansen, Philipp ; Holland-Letz, Tim ; Schilling, Bastian ; von Kalle, Christof ; Fröhling, Stefan ; Gaiser, Maria R. ; Hartmann, Daniela ; Gesierich, Anja ; Kähler, Katharina C. ; Wehkamp, Ulrike ; Karoglan, Ante ; Bär, Claudia ; Brinker, Titus J. ; Schmitt, Laurenz ; Peitsch, Wiebke K. ; Hoffmann, Friederike ; Becker, Jürgen C. ; Drusio, Christina ; Jansen, Philipp ; Klode, Joachim ; Lodde, Georg ; Sammet, Stefanie ; Schadendorf, Dirk

; Sondermann, Wiebke ; Ugurel, Selma ; Zader, Jeannine ; Enk, Alexander ; Salzmann, Martin ; Schäfer, Sarah ; Schäkel, Knut ; Winkler, Julia ; Wölbing, Priscilla ; Asper, Hiba ; Bohne, Ann-Sophie ; Brown, Victoria ; Burba, Bianca ; Deffaa, Sophia ; Dietrich, Cecilia ; Dietrich, Matthias ; Drerup, Katharina Antonia ; Egberts, Friederike ; Erkens, Anna-Sophie ; Greven, Salim ; Harde, Viola ; Jost, Marion ; Kaeding, Merit ; Kosova, Katharina ; Lischner, Stephan ; Maagk, Maria ; Messinger, Anna Laetitia ; Metzner, Malte ; Motamedi, Rogina ; Rosenthal, Ann-Christine ; Seidl, Ulrich ; Stemmermann, Jana ; Torz, Kaspar ; Velez, Juliana Giraldo ; Haiduk, Jennifer ; Alter, Mareike ; Bär, Claudia ; Bergenthal, Paul ; Gerlach, Anne ; Holtorf, Christian ; Karoglan, Ante ; Kindermann, Sophie ; Kraas, Luise ; Felcht, Moritz ; Gaiser, Maria R. ; Klemke, Claus-Detlev ; Kurzen, Hjalmar ; Leibing, Thomas ; Müller, Verena ; Reinhard, Raphael R. ; Utikal, Jochen ; Winter, Franziska ; Berking, Carola ; Eicher, Laurie ; Hartmann, Daniela ; Heppt, Markus ; Kilian, Katharina ; Krammer, Sebastian ; Lill, Diana ; Niesert, Anne-Charlotte ; Oppel, Eva ; Sattler, Elke ; Senner, Sonja ; Wallmichrath, Jens ; Wolff, Hans ; Giner, Tina ; Glutsch, Valerie ; Kerstan, Andreas ; Presser, Dagmar ; Schrüfer, Philipp ; Schummer, Patrick ; Stolze, Ina ; Weber, Judith ; Drexler, Konstantin ; Haferkamp, Sebastian ; Mickler, Marion ; Stauner, Camila Toledo ; Thiem, Alexander

Alternative links to fulltext:DOI Publisher

Details

Indexed in

Export bibliographical data

Abstract

Abstract

Background: Recently, convolutional neural networks (CNNs) systematically outperformed dermatologists in distinguishing dermoscopic melanoma and nevi images. However, such a binary classification does not reflect the clinical reality of skin cancer screenings in which multiple diagnoses need to be taken into account. Methods: Using 11,444 dermoscopic images, which covered dermatologic diagnoses comprising the majority of commonly pigmented skin lesions commonly faced in skin cancer screenings, a CNN was trained through novel deep learning techniques. A test set of 300 biopsy-verified images was used to compare the classifier's performance with that of 112 dermatologists from 13 German university hospitals. The primary end-point was the correct classification of the different lesions into benign and malignant. The secondary end-point was the correct classification of the images into one of the five diagnostic categories. Findings: Sensitivity and specificity of dermatologists for the primary end-point were 74.4% (95% confidence interval [CI]: 67.0-81.8%) and 59.8% (95% CI: 49.8-69.8%), respectively. At equal sensitivity, the algorithm achieved a specificity of 91.3% (95% CI: 85.5-97.1%). For the secondary end-point, the mean sensitivity and specificity of the dermatologists were at 56.5% (95% CI: 42.8-70.2%) and 89.2% (95% CI: 85.0-93.3%), respectively. At equal sensitivity, the algorithm achieved a specificity of 98.8%. Two-sided McNemar tests revealed significance for the primary end-point (p < 0.001). For the secondary end-point, outperformance (p < 0.001) was achieved except for basal cell carcinoma (on-par performance). Interpretation: Our findings show that automated classification of dermoscopic melanoma and nevi images is extendable to a multiclass classification problem, thus better reflecting clinical differential diagnoses, while still outperforming dermatologists at a significant level (p < 0.001). (C) 2019 The Author(s). Published by Elsevier Ltd.

Owner only: item control page

Altmetric

Alternative Statistics (altmetrics)

More literature (via CORE)

Details

Indexed in

Export bibliographical data

Abstract

Abstract

Alternative Statistics (altmetrics)

More literature (via CORE)

University Library