by ecancer reporter Janet Fricker

A computer trained to classify skin cancers has been shown to perform on a level with experienced dermatologists in diagnosing keratinocyte carcinomas and malignant melanomas, reports a study in Nature.

Skin cancer, the most common human malignancy, tends to be diagnosed visually, and then confirmed with follow-up biopsies and histological tests.

Early detection of melanoma is critical since estimated five-year survival rates drop from over 99% if detected in its earliest stages to about 14% if detected in its latest stages.

Previous attempts to develop automated classification systems have been difficult due to wide variation in the appearance of lesions.

In the study Andrew Esteva, and colleagues (Stanford University, California) used an algorithmic technique known as deep convolutional neural networks (CNNs) to train a computer to develop artificial intelligence in pattern recognition.

For training, the team used a set of 129,450 images of skin lesions and the names of the conditions each represented.

Altogether, 2,032 different skin diseases were included in the set.

The investigators then presented a set of previously unseen digital images of skin lesions to the trained computer and to 21 board certified dermatologists, with results verified through biopsy testing.

The team reports that the computer algorithms achieved performance on par with all tested experts when distinguishing keratinocyte carcinomas from benign seborrheic keratoses and malignant melanomas from benign nevi.

“This fast, scalable method is deployable on mobile devices and holds the potential for substantial clinical impact, including broadening the scope of primary care practice and augmenting clinical decision-making for dermatology specialists,” write the authors.

Notably, they add, the training set of images was about 100 times larger than any reported previously.

The ability to classify skin lesion images with the accuracy of a board-certified dermatologist has the potential to profoundly expand access to vital medical care.

Furthermore, if installed on mobile phones it could offer low-cost universal access to vital diagnostic care.

In an accompanying News & Views article, Sancy Leachman (Oregon Health and Science University, Portland),and Glenn Merlino (The National Cancer Institute, Bethesda, Maryland) write, “An obvious potential societal benefit of artificial intelligence in diagnostic technology would be improved access to high-quality health care.”

A Smartphone app involving this technology, they add, might enable effective, easy and low-cost medical assessments of more individuals than currently possible with existing medical care systems.

However, there could be adverse medical consequences, they caution, including medical staff becoming mere technicians responding to the machine’s diagnostic decisions.

It should not be long, they conclude, before there is a real Smartphone equivalent of the Star Trek tricorder.

This was the portable diagnostic device used by Dr Leonard McCoy in the1960s science-fiction television series to assess the medical condition of Enterprise crew members.

Reference:

Esteva A, Kuprel B, Novoa R, et al. Dermatologist – level classification of skin cancer with deep neural networks. Nature.