Brinker, Titus J. and Hekler, Achim and Enk, Alexander H. and Klode, Joachim and Hauschild, Axel and Berking, Carola and Schilling, Bastian and Haferkamp, Sebastian and Schadendorf, Dirk and Froehling, Stefan and Utikal, Jochen S. and von Kalle, Christof and Ludwig-Peitsch, Wiebke and Sirokay, Judith and Heinzerling, Lucie and Albrecht, Magarete and Baratella, Katharina and Bischof, Lena and Chorti, Eleftheria and Dith, Anna and Drusio, Christina and Giese, Nina and Gratsias, Emmanouil and Griewank, Klaus and Hallasch, Sandra and Hanhart, Zdenka and Herz, Saskia and Hohaus, Katja and Jansen, Philipp and Jockenhoefer, Finja and Kanaki, Theodora and Knispel, Sarah and Leonhard, Katja and Martaki, Anna and Matei, Liliana and Matull, Johanna and Olischewski, Alexandra and Petri, Maximilian and Placke, Jan-Malte and Raub, Simon and Salva, Katrin and Schlott, Swantje and Sody, Elsa and Steingrube, Nadine and Stoffels, Ingo and Ugurel, Selma and Sondermann, Wiebke and Zaremba, Anne and Gebhardt, Christoffer and Booken, Nina and Christolouka, Maria and Buder-Bakhaya, Kristina and Bokor-Billmann, Therezia and Enk, Alexander and Gholam, Patrick and Haenssle, Holger and Salzmann, Martin and Schaefer, Sarah and Schaekel, Knut and Schank, Timo and Bohne, Ann-Sophie and Deffaa, Sophia and Drerup, Katharina and Egberts, Friederike and Erkens, Anna-Sophie and Ewald, Benjamin and Falkvoll, Sandra and Gerdes, Sascha and Harde, Viola and Hauschild, Axel and Jost, Marion and Kosova, Katja and Messinger, Laetitia and Metzner, Malte and Morrison, Kirsten and Motamedi, Rogina and Pinczker, Anja and Rosenthal, Anne and Scheller, Natalie and Schwarz, Thomas and Stoelzl, Dora and Thielking, Federieke and Tomaschewski, Elena and Wehkamp, Ulrike and Weichenthal, Michael and Wiedow, Oliver and Baer, Claudia Maria and Bender-Saebelkampf, Sophia and Horbruegger, Marc and Karoglan, Ante and Kraas, Luise and Faulhaber, Joerg and Geraud, Cyrill and Guo, Ze and Koch, Philipp and Linke, Miriam and Maurier, Nolwenn and Mueller, Verena and Thomas, Benjamin and Utikal, Jochen Sven and Alamri, Ali Saeed M. and Baczako, Andrea and Berking, Carola and Betke, Matthias and Haas, Carolin and Hartmann, Daniela and Heppt, Markus V. and Kilian, Katharina and Krammer, Sebastian and Lapczynski, Natalie Lidia and Mastnik, Sebastian and Nasifoglu, Suzan and Ruini, Cristel and Sattler, Elke and Schlaak, Max and Wolff, Hans and Achatz, Birgit and Bergbreiter, Astrid and Drexler, Konstantin and Ettinger, Monika and Haferkamp, Sebastian and Halupczok, Anna and Hegemann, Marie and Dinauer, Verena and Maagk, Maria and Mickler, Marion and Philipp, Biance and Wilm, Anna and Wittmann, Constanze and Gesierich, Anja and Glutsch, Valerie and Kahlert, Katrin and Kerstan, Andreas and Schilling, Bastian and Schruefer, Philipp (2019) A convolutional neural network trained with dermoscopic images performed on par with 145 dermatologists in a clinical melanoma image classification task. EUROPEAN JOURNAL OF CANCER, 111. pp. 148-154. ISSN 0959-8049, 1879-0852
Full text not available from this repository. (Request a copy)Abstract
Background: Recent studies have demonstrated the use of convolutional neural networks (CNNs) to classify images of melanoma with accuracies comparable to those achieved by board-certified dermatologists. However, the performance of a CNN exclusively trained with dermoscopic images in a clinical image classification task in direct competition with a large number of dermatologists has not been measured to date. This study compares the performance of a convolutional neuronal network trained with dermoscopic images exclusively for identifying melanoma in clinical photographs with the manual grading of the same images by dermatologists. Methods: We compared automatic digital melanoma classification with the performance of 145 dermatologists of 12 German university hospitals. We used methods from enhanced deep learning to train a CNN with 12,378 open-source dermoscopic images. We used 100 clinical images to compare the performance of the CNN to that of the dermatologists. Dermatologists were compared with the deep neural network in terms of sensitivity, specificity and receiver operating characteristics. Findings: The mean sensitivity and specificity achieved by the dermatologists with clinical images was 89.4% (range: 55.0%-100%) and 64.4% (range: 22.5%-92.5%). At the same sensitivity, the CNN exhibited a mean specificity of 68.2% (range 47.5%-86.25%). Among the dermatologists, the attendings showed the highest mean sensitivity of 92.8% at a mean specificity of 57.7%. With the same high sensitivity of 92.8%, the CNN had a mean specificity of 61.1%. Interpretation: For the first time, dermatologist-level image classification was achieved on a clinical image classification task without training on clinical images. The CNN had a smaller variance of results indicating a higher robustness of computer vision compared with human assessment for dermatologic image classification tasks. (C) 2019 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
| Item Type: | Article |
|---|---|
| Uncontrolled Keywords: | DIAGNOSIS; ALGORITHMS; Melanoma; Artificial intelligence; Diagnostics; Skin cancer |
| Subjects: | 600 Technology > 610 Medical sciences Medicine |
| Divisions: | Medicine > Lehrstuhl für Dermatologie und Venerologie |
| Depositing User: | Dr. Gernot Deinzer |
| Date Deposited: | 15 Apr 2020 07:34 |
| Last Modified: | 15 Apr 2020 07:34 |
| URI: | https://pred.uni-regensburg.de/id/eprint/27300 |
Actions (login required)
 |
View Item |
