In the last couple of decades, human pathologists have been the ones giving the cancer diagnosis through a time-intensive work. Researchers are now trying to train machines to do the same with the help of artificial intelligence. A new research that was published in Nature Medicine shows how scientists from the New York University have re-instructed a Google deep learning algorithm to differentiate between two most common types of lung cancer. The accuracy was an astonishing 97%.
The AI that is able to identify tumors is the same technology that is used to identify faces, objects, animals in the pictures that can be uploaded to the online services that Google has. It has been used before in order to diagnose heart conditions and diabetic blindness. However, what happened recently is something that no pathologist was able to do before. The neural network from NYU was able to recognize the genetic mutations from each tumor, and just by using one single picture.
AI will definitely change the future of the medicine
Aristotelis Tsirigos, who is a lead author of the new study and a pathologist at the NYU School of Medicine, stated that “the real novelty” in this situation was not to just demonstrate that AI is as good as humans can be, but to also show that artificial intelligence can, in fact, give some insight – one which a human expert could not be able to provide.
In order to teach Google’ algorithm, Inception v3, how to differentiate between pictures of healthy and cancerous tissue, the researchers had to show it hundreds of thousands of images that were taken from a public library of patient tissue samples, The Cancer Genome Atlas.
However, this was just the first step. Next, once the algorithm knew how to choose the cancerous cells with an accuracy of 99%, the second step was to instruct it how to distinguish between two different types of lung cancer. Under the microscope, these two kinds of cancer seem very similar, but they are treated very differently. You can imagine that such a difference can actually determine if a patient will live or die. The most incredible thing in this research is that the algorithm learned how to identify genetic mutations by itself and not with the help of a human hand and that is truly a breakthrough.
Karen and her husband live on a plot of land in British Columbia. They aim to grow and raise a significant part of their food by maintaining a vegetable garden, keeping a flock of backyard chickens and foraging. They are also currently planning a move to a small cabin they hand built. Karen’s academic background in nutrition made her care deeply about real food and seek ways to obtain it. Thus sprung Anna’s interest in backyard gardening, chicken and goat keeping, recycling and self-sufficiency.