Journal Article

DKFZ-2019-01167

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png Deep learning outperformed 136 of 157 dermatologists in a head-to-head dermoscopic melanoma image classification task.

Brinker, T. J. (First author)DKFZ* ; Hekler, A.DKFZ* ; Enk, A. H. ; Klode, J. ; Hauschild, A. ; Berking, C. ; Schilling, B. ; Haferkamp, S. ; Schadendorf, D. ; Holland-Letz, T.DKFZ* ; Utikal, J. S. (Last author)DKFZ* ; von Kalle, C. (Last author)DKFZ* ; Collaborators (Collaboration Author) ; Ludwig-Peitsch, W. ; Sirokay, J. ; Heinzerling, L. ; Albrecht, M. ; Baratella, K. ; Bischof, L. ; Chorti, E. ; Dith, A. ; Drusio, C. ; Giese, N. ; Gratsias, E. ; Griewank, K. ; Hallasch, S. ; Hanhart, Z. ; Herz, S. ; Hohaus, K. ; Jansen, P. ; Jockenhöfer, F. ; Kanaki, T. ; Knispel, S. ; Leonhard, K. ; Martaki, A. ; Matei, L. ; Matull, J. ; Olischewski, A. ; Petri, M. ; Placke, J.-M. ; Raub, S. ; Salva, K. ; Schlott, S. ; Sody, E. ; Steingrube, N. ; Stoffels, I. ; Ugurel, S. ; Zaremba, A. ; Gebhardt, C. ; Booken, N. ; Christolouka, M. ; Buder-Bakhaya, K. ; Bokor-Billmann, T. ; Enk, A. ; Gholam, P. ; Hänßle, H. ; Salzmann, M. ; Schäfer, S. ; Schäkel, K. ; Schank, T. ; Bohne, A.-S. ; Deffaa, S. ; Drerup, K. ; Egberts, F. ; Erkens, A.-S. ; Ewald, B. ; Falkvoll, S. ; Gerdes, S. ; Harde, V. ; Hauschild, A. ; Jost, M. ; Kosova, K. ; Messinger, L. ; Metzner, M. ; Morrison, K. ; Motamedi, R. ; Pinczker, A. ; Rosenthal, A. ; Scheller, N. ; Schwarz, T. ; Stölzl, D. ; Thielking, F. ; Tomaschewski, E. ; Wehkamp, U. ; Weichenthal, M. ; Wiedow, O. ; Bär, C. M. ; Bender-Säbelkampf, S. ; Horbrügger, M. ; Karoglan, A. ; Kraas, L. ; Faulhaber, J. ; Geraud, C. ; Guo, Z. ; Koch, P. ; Linke, M. ; Maurier, N. ; Müller, V. ; Thomas, B. ; Utikal, J. S.DKFZ* ; Alamri, A. S. M. ; Baczako, A. ; Berking, C. ; Betke, M. ; Haas, C. ; Hartmann, D. ; Heppt, M. V. ; Kilian, K. ; Krammer, S. ; Lapczynski, N. L. ; Mastnik, S. ; Nasifoglu, S. ; Ruini, C. ; Sattler, E. ; Schlaak, M. ; Wolff, H. ; Achatz, B. ; Bergbreiter, A. ; Drexler, K. ; Ettinger, M. ; Haferkamp, S. ; Halupczok, A. ; Hegemann, M. ; Dinauer, V. ; Maagk, M. ; Mickler, M. ; Philipp, B. ; Wilm, A. ; Wittmann, C. ; Gesierich, A. ; Glutsch, V. ; Kahlert, K. ; Kerstan, A. ; Schilling, B. ; Schrüfer, P.

2019
Elsevier Amsterdam [u.a.]

This record in other databases:  

Please use a persistent id in citations: doi:10.1016/j.ejca.2019.04.001

Abstract: Recent studies have successfully demonstrated the use of deep-learning algorithms for dermatologist-level classification of suspicious lesions by the use of excessive proprietary image databases and limited numbers of dermatologists. For the first time, the performance of a deep-learning algorithm trained by open-source images exclusively is compared to a large number of dermatologists covering all levels within the clinical hierarchy.We used methods from enhanced deep learning to train a convolutional neural network (CNN) with 12,378 open-source dermoscopic images. We used 100 images to compare the performance of the CNN to that of the 157 dermatologists from 12 university hospitals in Germany. Outperformance of dermatologists by the deep neural network was measured in terms of sensitivity, specificity and receiver operating characteristics.The mean sensitivity and specificity achieved by the dermatologists with dermoscopic images was 74.1% (range 40.0%-100%) and 60% (range 21.3%-91.3%), respectively. At a mean sensitivity of 74.1%, the CNN exhibited a mean specificity of 86.5% (range 70.8%-91.3%). At a mean specificity of 60%, a mean sensitivity of 87.5% (range 80%-95%) was achieved by our algorithm. Among the dermatologists, the chief physicians showed the highest mean specificity of 69.2% at a mean sensitivity of 73.3%. With the same high specificity of 69.2%, the CNN had a mean sensitivity of 84.5%.A CNN trained by open-source images exclusively outperformed 136 of the 157 dermatologists and all the different levels of experience (from junior to chief physicians) in terms of average specificity and sensitivity.

---

Contributing Institute(s):

1. Translationale Medizinische Onkolologie (B340)
2. Biostatistik (C060)
3. KKE Dermatoonkologie (A370)

Research Program(s):

1. 312 - Functional and structural genomics (POF3-312) (POF3-312)

Appears in the scientific report 2019

---

Database coverage:
; BIOSIS Previews ; Clarivate Analytics Master Journal List ; Current Contents - Clinical Medicine ; Current Contents - Life Sciences ; Ebsco Academic Search ; IF >= 5 ; JCR ; NCBI Molecular Biology Database ; Nationallizenz ; SCOPUS ; Science Citation Index ; Science Citation Index Expanded ; Web of Science Core Collection

---