A team of researchers from the US, Germany and France has taught an AI system to distinguish benign skin lesions from dangerous ones, reports The Guardian.
The team used some 100,000 images to get the AI's deep learning convolutional neural network (CNN) up-to-scratch. Once trained, it was tested against 58 dermatologists from across 17 countries, who were shown benign moles and malignant melanomas.
The AI accurately identified 95 per cent of skin cancers from the images, compared to the 86.6 per cent accuracy of the human dermatologists. It's not a huge win for the AI, but it still proved its better than its fleshy counterparts, which had experience that ranged from less than two years to more than five.
It also showed how the rapid the advancement of AI tech can be, given it was only 14 months ago that scientists at Stamford University found that AI was just as good at spotting skin cancer as humans are.
"The CNN missed fewer melanomas, meaning it had a higher sensitivity than the dermatologists," said Holger Haenssle, the first author of the researcher's paper and an academic at the University of Heidelberg, noted The Guardian.
Such an AI system could, according to the researchers, result in "less unnecessary surgery" as it misdiagnosed fewer benign moles as malignant melanoma. This would suggest that human dermatologists tend to over-diagnose malignant melanomas, playing it safe rather than risk passing of a dangerous mole as benign.
It's not likely that AI will replace the efforts of dermatologists any time soon, as a thorough clinical examination is a safer bet than a clever system adept at image recognition. But such a system could help make life easier for clinicians under pressure and act as the first examination of a suspect mole to provide more information for a close clinician's inspection.
AI is being used for some potentially scary and/or useful stuff around the world, but the overwhelming majority of the population seem to think that it'll help the world rather than, I don't know, kill us all.
New cable will connect Virginia to France
Loon's balloons will bring the internet to remote areas of the country
New clues into the biosphere on Earth in the lead up to the emergence of animal life
Planetary collision might shed light on the chaotic processes behind a star's early development