Computer System Reported To Have Found Examples Of Melanomas Much Quicker Than A Medically-Trained Dermatologist

There is no question we continue to be surrounded by computers, and it seems with each passing day they get more and more intelligent. In recent news, an artificial intelligence app was able to correctly diagnose cases of skin cancer 95% of the time, which outpaced a panel of human dermatologists, who were only able to correctly diagnose skin cancer 86% of the time. Of course, teaching this artificial intelligence app to do the job was a daunting task in of itself. A computer engineering team in Germany taught this program, called the convolutional neural network (CNN), to correctly diagnose skin cancer more often than doctors by showing it over 100,000 images of both malignant and benign skin lesions. Of course, that is when they proceeded to test their app against 58 medically trained dermatologist in many different countries across the globe.

The sample size of the dermatologists showed that half of them were at an expert level, meaning they had 5 years of experience or more. 19% of the dermatologist had moderate experience from 2 to 5 years, and 29% of them were complete novices with just less than a couple years experience diagnosing melanomas. The machine missed fewer melanomas because it was designed by the computer engineers to be more sensitive to various stimuli. However, that isn’t the end of the story.

The melanomas on the fingers, toes, and scalp were difficult for the artificial intelligence app to properly photograph. Moreover, physicians also did better when they had a documentation of all patient histories. Either way, many medical personnel are excited about the capabilities of this artificial intelligence system. They realize it will be faster and easier to diagnose many forms of skin cancer and save more lives in the future. That is definitely something each and every one of us can get behind.

Leave a Reply

Your email address will not be published. Required fields are marked *