Melanoma is the most dangerous type of skin cancer but is highly treatable in the early stages and much harder to eradicate later on.
This makes the early detection pivotal but, often, both the large number of patients and the symptomatic proximity of many skin diseases concur to leave melanoma overlooked.
In the last years, several automated, low cost systems have been developed - able to assess a patient’s risk of melanoma using images of his/hers skin lesions captured using a standard digital camera. The enabling factor for these system is both the use of efficient deep learning algorithms, as the employment of large image datasets.
In this regard, there are currently many datasets published on internet databases. One of the most important ones is Dermnet.com database, accounting almost 22.500 images divided in 23 different classes. Despite its images being repeatedly employed in peer reviewed publications, they rightly ask a prior written consent for their use “for any purpose including but not limited to research, commercial, personal, or non-commercial”. Notwithstanding the many emails I sent using their internal licensing system by now (the first one dates back to May 2016), I never received any response neither positive nor negative. This is becoming a pretty big drawback since me and my group are unable to proceed with our study on melanoma diagnosis based on a computer vision approach.
I do not know how did the authors solve this problem and, I suppose, someone may have simply ignored the warning since they had not commercial purposes. My purpose is exclusively non-profit, but I refuse to proceed in unlawful ways. Does anybody have some suggestions or a contact to suggest in order to go on with this problem? Otherwise, did somebody face this same problem in his experience?
- Badges
- Science topic