A team of RECOD researchers participated for the second time on the melanoma classification task at the melanoma diagnostic challenge promoted by the International Skin Imaging Collaboration. This year the number of participants grew 6 times regarding last edition, reaching 77 participants — our team got 6th place. Competitors all around the world submitted their approaches, including private companies focused on smart skin analytics.
RECOD also participated on the tasks of lesion segmentation and lesion attribute detection. RECOD’s participation in those tasks is detailed in a technical report (submitted before the official ranking was announced).
The global results will be discussed at the upcoming Medical Image Computing & Computer Assisted Intervention (MICCAI 2018), where RECOD team will also present two papers (out of ten selected) at the ISIC Skin Image Analysis Workshop, regarding new approaches for dermatologic dataset extension: techniques for data augmentation and synthetic image generation with GANs.
We gave a preview of our synthetic image generation with GANs at the 2nd International Educational Symposium of the Melanoma World on Rio de Janeiro (August, 2018), were it got the 2nd place at the best poster awards.
The melanoma team is composed by professors Eduardo Valle and Sandra Avila, Ph.D. student Michel Fornaciali, and M.Sc. students Alceu Bissoto, Fábio Perez and Vinícius Ribeiro, all RECOD members.