English

Summary

The article "Deep learning algorithm does as well as dermatologists in identifying skin cancer" by Taylor Kubota gives a review of a new technique using convolutional neural networks to diagnose melanoma by image. According to the article, computer scientists at Stanford created a dataset consisting of about 130 000 image of skin diseases and managed to train a network to successfully detect cancer. The author states, that it is essential to diagnose melanoma in early stages, therefore a easy way of doing this could be very important. The article gives information on what model did the researches choose, how did they create their database and how did they evaluate their network. The final performance of the algorithm is stated to be nearly the same as the human dermatologists' one. In the end, author says that the scientists want to make their algorithm available on smart-phones. In their opinion, it should not be too hard to do, but some more experiments should be carried out to prove that the model works correctly in clinical practice.