Scientists will soon make it easier for people to identify skin cancer. The researchers at Stanford University created a project that proposes to intertwine deep learning neural networks and dermatology.
The computer scientists at Stanford used a total of about 130,000 images of different skin diseases and created an algorithm for artificial intelligence that can identify and diagnose skin conditions with an incredible accuracy matched only by trained professionals.
One of the main authors of the research, Andre Esteva, talked about the project. He explained that they trained the AI to look at images of skin diseases and identify if they are malignant or not. The results showed that the AI matched the performance of more than 21 certified dermatologists.
The diagnostics that the machine was able to identify were keratinocyte carcinomas, which represent the most common type of skin cancer, and melanoma, representing the deadliest type of skin cancer. The machine was also able to identify melanoma while using dermoscopy.
The researchers did not create a new neural network, but they used a previously created one. It was originally designed by Google to differentiate images, indeed, but it is not what you would expect. The neural network was meant to tell cats from dogs. However, the researchers found that it was pretty good at distinguishing more than 200 types of dogs. Brett Kuprel, one of the co-authors, described the ability of the machine as “superhuman”.
This impressive ability inspired the Stanford researchers to use the AI technology for a more useful purpose. Then the brilliant idea of identifying skin cancer came to them. Unlike the expectations, none of the researchers had any background in dermatology, therefore the machine did not benefit from any expert insight into the field.
However, AI-skeptics should not worry that this machine is meant to replace humans. It might prove as a useful tool that can help dermatologists in their diagnosis. It may work as a kind of an X-ray machine that can offer a second opinion on the skin condition.
There are still many steps until such a machine gets functional. It requires an FDA approval and still some improvements until it is good to use. Regardless of this, the idea is still a revolutionary one and people are now closer to more accurate diagnoses of skin cancer.
Image Source: Pixabay