A team of scientists developed an artificial intelligence algorithm that was able to accurately detect and diagnose skin cancer based just on images and deep learning tools.
Skin cancer is not the most easily identifiable disease. It can often get confused or overlooked. It depends on a visual exam. This examines the potential lesion. This initial look can be followed by a dermoscopic analysis. If necessary, the patient will also have to offer a biopsy sample. And only then will they receive a histologic confirmation. The process can take some time.
Now, scientists have developed a new type of examination. A team of Stanford University researchers developed an artificial intelligence powered algorithm.
Their studies on the matter were released earlier this week. A paper was published in the Nature journal. It became available online on January 25 and titled as follows. “Dermatologist-level classification of skin cancer with deep neural networks”. Research was led by Andre Esteva.
Together with his team, Esteva developed a new, artificial intelligence powered computational method. Esteva released a statement. According to it, he hopes that the algorithm will have a widespread use. Doctors could be its first potential utilizers. It may help them with the early detection of skin cancer. And it may also help determine if a tumor is benign or malign.
Esteva offered some details about the algorithm. The artificial intelligence system is based on deep learning. This means that the system will keep learning as it is fed data.
The current algorithm is based on an existing tool. A Google image recognition system was used as a base. Then, the researchers started training the algorithm. They used more than 13,000 images of skin lesions. These represented a total of 2,000 diseases.
Their artificial intelligence system was trained to distinguish them. And also place them in the accurate category. As such, it can detect cancer lesions. And help diagnose skin cancer. But at the same time, it can also point out if a lesion is benign.
Their artificial intelligence or AI algorithm was then put to the test. And the early test results were quite outstanding. Reportedly, the system had an accuracy rate of “at least” 91 percent. This was registered in comparison to the flesh and blood doctors.
Still, the AI algorithm still requires some more work. It has to be further refined. The researchers will continue training it. They want to prevent any potential misdiagnosis. And the artificial intelligence system will also have to be vigorously analyzed.
Its efficiency will have to be put to the test. As well as its working powers. The scientists will also be working on trying to prevent glitches.
Still, this artificial intelligence algorithm could become an invaluable tool. It remains to be seen if it will be released for a widespread use. And when that might happen. But the system could work wonders. It could come to help doctors put a diagnosis. And an early one at that, without the need for several tests.
If proven effective, the system could be further developed. It could even potentially be transformed into an application. This could supposedly allow any user to do a quick test. As such, theoretically, anyone could check out any suspicious looking lesion.
Nonetheless, there is still quite a bit of road left till we’ll establish such a technology. But the current artificial intelligence could be a step in that direction.
Image Source: Wikimedia