% IMPORTANT: The following is UTF-8 encoded.  This means that in the presence
% of non-ASCII characters, it will not work with BibTeX 0.99 or older.
% Instead, you should use an up-to-date BibTeX implementation like “bibtex8” or
% “biber”.

@ARTICLE{Brinker:143587,
      author       = {T. J. Brinker$^*$ and A. Hekler$^*$ and A. H. Enk and J.
                      Klode and A. Hauschild and C. Berking and B. Schilling and
                      S. Haferkamp and D. Schadendorf and T. Holland-Letz$^*$ and
                      J. S. Utikal$^*$ and C. von Kalle$^*$ and W. Ludwig-Peitsch
                      and J. Sirokay and L. Heinzerling and M. Albrecht and K.
                      Baratella and L. Bischof and E. Chorti and A. Dith and C.
                      Drusio and N. Giese and E. Gratsias and K. Griewank and S.
                      Hallasch and Z. Hanhart and S. Herz and K. Hohaus and P.
                      Jansen and F. Jockenhöfer and T. Kanaki and S. Knispel and
                      K. Leonhard and A. Martaki and L. Matei and J. Matull and A.
                      Olischewski and M. Petri and J.-M. Placke and S. Raub and K.
                      Salva and S. Schlott and E. Sody and N. Steingrube and I.
                      Stoffels and S. Ugurel and A. Zaremba and C. Gebhardt and N.
                      Booken and M. Christolouka and K. Buder-Bakhaya and T.
                      Bokor-Billmann and A. Enk and P. Gholam and H. Hänßle and
                      M. Salzmann and S. Schäfer and K. Schäkel and T. Schank
                      and A.-S. Bohne and S. Deffaa and K. Drerup and F. Egberts
                      and A.-S. Erkens and B. Ewald and S. Falkvoll and S. Gerdes
                      and V. Harde and A. Hauschild and M. Jost and K. Kosova and
                      L. Messinger and M. Metzner and K. Morrison and R. Motamedi
                      and A. Pinczker and A. Rosenthal and N. Scheller and T.
                      Schwarz and D. Stölzl and F. Thielking and E. Tomaschewski
                      and U. Wehkamp and M. Weichenthal and O. Wiedow and C. M.
                      Bär and S. Bender-Säbelkampf and M. Horbrügger and A.
                      Karoglan and L. Kraas and J. Faulhaber and C. Geraud and Z.
                      Guo and P. Koch and M. Linke and N. Maurier and V. Müller
                      and B. Thomas and J. S. Utikal$^*$ and A. S. M. Alamri and
                      A. Baczako and C. Berking and M. Betke and C. Haas and D.
                      Hartmann and M. V. Heppt and K. Kilian and S. Krammer and N.
                      L. Lapczynski and S. Mastnik and S. Nasifoglu and C. Ruini
                      and E. Sattler and M. Schlaak and H. Wolff and B. Achatz and
                      A. Bergbreiter and K. Drexler and M. Ettinger and S.
                      Haferkamp and A. Halupczok and M. Hegemann and V. Dinauer
                      and M. Maagk and M. Mickler and B. Philipp and A. Wilm and
                      C. Wittmann and A. Gesierich and V. Glutsch and K. Kahlert
                      and A. Kerstan and B. Schilling and P. Schrüfer},
      collaboration = {Collaborators},
      title        = {{D}eep learning outperformed 136 of 157 dermatologists in a
                      head-to-head dermoscopic melanoma image classification
                      task.},
      journal      = {European journal of cancer},
      volume       = {113},
      issn         = {0959-8049},
      address      = {Amsterdam [u.a.]},
      publisher    = {Elsevier},
      reportid     = {DKFZ-2019-01167},
      pages        = {47 - 54},
      year         = {2019},
      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.},
      cin          = {B340 / C060 / A370},
      ddc          = {610},
      cid          = {I:(DE-He78)B340-20160331 / I:(DE-He78)C060-20160331 /
                      I:(DE-He78)A370-20160331},
      pnm          = {312 - Functional and structural genomics (POF3-312)},
      pid          = {G:(DE-HGF)POF3-312},
      typ          = {PUB:(DE-HGF)16},
      pubmed       = {pmid:30981091},
      doi          = {10.1016/j.ejca.2019.04.001},
      url          = {https://inrepo02.dkfz.de/record/143587},
}