We are searching data for your request:
Upon completion, a link will appear to access the found materials.
In the future, reliable skin cancer diagnosis will be possible using smartphones
With skin cancer, early detection is important for optimal treatment of the disease. Researchers have now developed an artificial intelligence system that can detect the telltale signs of skin cancer as reliably as human doctors can. This new technology should work on every smartphone in the future.
Scientists at Stanford University found that skin cancer can be diagnosed more quickly and easily with the help of artificial intelligence. In the future, people might be able to self-diagnose using their smartphones. The system works just as reliably as human doctors. The doctors published the results of their study in the journal "Nature".
In the future, those affected may be able to screen themselves
The potential of the newly developed technology is great. If the system becomes more refined and portable, many people could screen themselves with minimal costs, the experts explain. This would save those affected from having to see a doctor and wait for them to confirm the suspicion.
Algorithm learns from existing data
The key to the success of the new technology is an algorithm that can learn from existing data, the researchers say. The system could use information and images from a database on skin cancer. "We have developed a very comprehensive learning algorithm that learns from existing data," says Andre Esteva of Stanford University.
System uses Google's algorithm
To convey its intelligence to the system, the researchers trained it with the help of 129,450 close-ups of skin lesions. These covered more than 2,000 different diseases and also provided an extensive database with examples. The team next used an algorithm developed by Google to recognize the difference between cats and dogs in pictures. The algorithm was then adjusted to detect the difference between different skin spots.
Artificial intelligence is just as successful as human doctors
The scientists then tested their system against 21 qualified dermatologists. The experts were shown 376 images of skin lesions. Then they were asked to rate the pictures. Based on the assessment, the doctors should then refer the patients for further analysis or remove the suspicion of skin cancer, the researchers explain. The scientists emphasize that the system with its artificial intelligence was able to achieve the success rate of the professionals.
System performance is impressive
The technology was of course not designed to replace doctors, the authors say. It was designed to give people easier access to the first two stages of screening before the help of experts, the researchers said. Recognizing the difference between a fatal lesion and a benign one is no easy task. This makes the system's performance even more impressive.
The aim is to operate the program via a telephone app
Before the device can be released to the public, it must be ensured that it does not create false ratings. Clinical trials should therefore help to improve the system even further. The doctors will add that the program can later be used via a telephone app.
Almost everyone has a “supercomputer” with sensors
Smartphones are ubiquitous these days. So in the future almost everyone should be able to use the system. Everyone has a supercomputer in their pocket, so to speak, which is equipped with a number of sensors, including a camera, says Esteva.
Assessment of our health from home?
We are now seeing numerous programs and applications that are driven by the intuitive way of thinking of artificial intelligence. These programs also run on our smartphones. This opens up cheap and easy ways to assess our health at home.
Early detection is crucial
As with many other diseases, early detection of skin cancer is crucial. If cancer is detected early, its 10-year survival rate is around 95 percent. However, this rate drops to ten to 15 percent, if the cancer has reached later stages before its treatment, the scientists explain. (as)