Published: Thu, May 31, 2018
Medicine | By Brett Sutton

AI Breakthrough Can Detect Skin Cancer Better Than Doctors


The scientists trained this artificial intelligence system to distinguish skin lesions and moles according to whether they were benign or malign, using more than 100,000 images. A CNN is an artificial neural network inspired by the biological processes at work when nerve cells (neurons) in the brain are connected to each other and respond to what the eye sees.

Researchers from Germany, the United States and France trained a deep learning convolutional neural network (CNN) to identify skin cancer by showing it over 100,000 images of malignant melanomas and benign moles.

The diagnostic credibility of the artificial intelligence system was compared to that of 58 dermatologists from 17 countries, including Greece.

"The CNN missed fewer melanomas, meaning it had a higher sensitivity than the dermatologists", said Holger Haenssle, first author of the study and a professor at the University of Heidelberg, Germany. This deep learning system is said to be capable of learning by "looking" at images and teaching itself what it has to know so it can gradually enhance its performance.

To be fair, the performance of the human dermatologists improved when they were provided more information about the patients - including age, gender, and position of the lesion - plus close-up images of skin lesions of the cases.


"Most dermatologists were outperformed by the CNN", said the research team. Worldwide, an estimated 232,000 new cases will be diagnosed in 2018, causing around 55,500 deaths. This is why this CNN is so impressive - it would be able to identify more cancers early on, thereby saving lives.

On average, the dermatologists correctly detected around 86.6 percent of melanomas while the CNN identified 95 percent of them. This would suggest that human dermatologists tend to over-diagnose malignant melanomas, playing it safe rather than risk passing of a risky mole as benign. But even so, the CNN, which was still working only with the images and was given no additional information, still outperformed the humans. Over 50% of the dermatologists were experts over 5 years of experience. Melanoma in some parts of the body, such as the fingers, toes, and scalp, are hard to image, and AI may have difficulty recognizing. For this reason, there is still no substitute for a thorough clinical examination performed by a trained human physician. "Still, there is much more work to be done to implement this exciting technology safely into routine clinical care", they continued.

According to the researchers, their deep learning network could one day help dermatologists in screening skin cancer and making the right decision to either biopsy a lesion or not.

Will Artificial Intelligence Take Over The World?

Like this: