Deep learning matches the performance of dermatologists at skin cancer classification
A research paper published by Stanford in Nature shows the Artificial Intelligence model trained to classify images of skin lesions as benign lesions or malignant skin cancers achieves the accuracy of board-certified dermatologists.
Benign and malignant skin lesions can be separated to a certain extent by looking at the margin, texture and color of lesions in skin by a dermatologist. Skin cancers such as melanoma is one of the most common neoplasm of skin, but in the meantime it also has a higher 5 year survival rate if detected earlier.
They pretrained a deep neural network at general object recognition, then fine-tune it on a dataset of ~130,000 skin lesion images comprised of over 2000 diseases to obtain the result.
This is a proof that Artificial Intelligence is going to redefine medicine.