University of Basel

Skinsifter for dermoscopy

Version 191010.1


Experimental classifier for skin lesions. Our deep learning model has been trained on ISIC2019 data and will classify into the following groups

Melanoma Melanocytic nevus
Basal cell carcinoma Squamous cell carcinoma
Actinic keratosis Benign keratosis*
Dermatofibroma Vascular lesion

* (Solar lentigo / seborrheic keratosis / lichen planus-like keratosis)


The model works better with images from patient older then 20 years old. For young persons, due to the growing of the patient, benign lesions use to look like malignant ones.

This algorithm is a scientific experiment intended for improving clinical machine learning and does not produce medical diagnoses. Using the website does not lead to a doctor-patient relationship and the result shall not be considered a medical advice. No claim is made that the result is more accurate than allocation by chance. By uploading a picture the User grants the University of Basel a (non-exclusive) royalty free right, unlimited in time and place to use the picture for the Research Purpose. All uploaded pictures are saved in a secure server at the University of Basel. As anonymized data the pictures may be used for future research. Neither University of Basel nor University Hospital of Basel assume any liability for the content or the results of the algorithm.

Created by Navarinilab at Department of Biomedical Engineering, University of Basel, Switzerland. Authors: Philippe Gottfrois, Ludovic Amruthalingam, Alexander A. Navarini