Verein für Hautkrebsforschung

Newsletter der Melanoma World Society, 1/2017

31-01-2017 Hautkrebsforschung

Skin Cancer Screening by Smartphones - Nature, January 2017

Claus Garbe, University Department of Dermatology, Tuebingen, Germany


Computed Diagnosis
Artificial intelligence can already solve many problems, and computer programs are able to beat the best chess players of the world. We even recently got to know that the best poker players have been beaten by computer algorithms. Why shouldn’t it not be possible to recognize skin cancers, particularly melanoma and also keratinocyte cancers, which are diagnosed by their visual characteristics either with the naked eye or with magnification by dermatoscopy.

Since two decades worked dermatological research groups, mainly in Germany, Austria and Italy, to develop diagnostic devices for melanoma. Several algorithms were developed and also sold on the dermatology market. These computer algorithms were integrated into devices for computer-dermatoscopy, and used by dermatologists in order to improve their diagnostic performance. The algorithms were based on logistic regression analysis or on neuronal networks utilizing image algebra. These computer programs developed already a high diagnostic performance comparable to that of dermatologists. However, these programs were mostly not trained to distinguish seborrheic keratoses from melanocytic lesions and they were not developed for the use of laypersons.

The New Breakthrough
Now seemingly a breakthrough in this field has been achieved. Sebastian Thrun, an adjunct professor in the Stanford Artificial Intelligence Laboratory developed with his coworkers André Esteva and Brett Kuprel in the Thrun laboratories together with dermatologists from the Pigmented Lesion and Melanoma Program at the Stanford Cancer Institute a new algorithm for the diagnosis of skin cancer.

It was essential for this project to build large collections of pictures from skin cancer, in the end, the Stanford project group succeeded to collect about 130,000 images of skin lesions representing different disease entities. They started with an algorithm developed by Google that was able to identify more than 1 million images from 1000 object categories. Among these, the program was able to differentiate pictures of cats from those of dogs. The algorithm had been tested in comparison to the diagnostic performance of 21 dermatologists.

Three main key diagnostic tasks were approached: The classification of keratinocyte carcinomas, the classification of melanomas by macroscopic and dermatoscopic pictures. In these three tasks the algorithm matched the performance of the dermatologists. Correct diagnoses were achieved in ~ 90% of the lesions. The results have been published in a research letter in Nature at January 25, 2017 (Esteva et al., 2017).

Perspectives
This is now the breakthrough for diagnostic algorithms, which can be transferred to devices with ability to analyze smartphone pictures. A number of questions have probably still to be solved, like a kind of standardization of photography, and also the marketing strategy for such an offer in healthcare. Nevertheless, there are few doubts that such a diagnostic algorithm may be a perfect precondition for new forms of skin cancer screening. What does this mean for dermatologists? Will they become needless? Probably we have to expect the contrary outcome: Laypersons using such computer programs will have urgent needs for explanation and management of their suspected skin cancer lesions!

In conclusion, let us support the development of such computer programs beating the diagnostic performance of dermatologists.

Esteva A, Kuprel B, Novoa RA, Ko J, Swetter SM, Blau HM, et al. Dermatologist-level
classification of skin cancer with deep neural networks. Nature 2017;542(7639):115-8.

Artikel als PDF

Hautkrebspreis 2017

Skin Cancer Award

Es freut uns sehr, Ihnen mitzuteilen, dass der Verein für Hautkrebsforschung in Zusammenarbeit mit der Firma Pierre Fabre 2012 einen Forschungspreis für junge Wissenschaftler im Bereich Dermato-Onkologie ins Leben gerufen hat.

Erfahren Sie mehr 

Newsletters

  • Neuigkeiten von der ESMO (European Society for Medical Oncology)

    Vom 8. – 12. September 2017 fand die Jahrestagung der European Society of Medical Oncology in Madrid statt. Für das Melanom der Haut gab es eine ganze Reihe von bahnbrechenden Ergebnissen.

    Written on Tuesday, 26 September 2017 00:00
Weitere

Unterstützen Sie uns!

Unterstützen Sie uns im Kampf gegen Hautkrebs

Mit einer Spende können Sie uns Im Kampf gegen Hautkrebs helfen!

Donatoren können selber entscheiden, welche Form von Unterstützung sie wählen möchten 

Erfahren Sie mehr