Skin cancer continues to be prevalent over the years. Millions of cases of non-melanoma skin cancers and hundreds of thousands of melanoma skin cancers happen each year. Melanoma is curable when detected early, but it could be harmful once it metastasizes to organs such as the lungs. Because of the continued harm to the ozone layer, the risk of a person getting skin cancer is higher than before.
However as technology continues to develop, people find ways to utilize it for different aspects of life. Artificial intelligence is also breaking through in people’s lives. It is used almost everywhere nowadays, especially in health care. Even though there are recent developments when it comes to the detection of health problems, it is still possible for these advancements not to cover every person’s needs.
At Stanford University, they developed an algorithm for detecting skin cancer wherein they gathered around 130,000 images of skin cancer and used those to train the algorithm in detecting skin cancer. They combined visual processing with deep learning, an AI which works just like neural networks in the brain. The technology used raw pixels with disease labels to further develop the algorithm’s efficiency.
Instead of starting from zero, the researchers used an existing algorithm by Google. The technology is already trained to detect millions of images from different categories.
To test the algorithm for detecting skin cancer, the researchers at Stanford used high-quality and biopsy-confirmed images from the University of Edinburgh and the International Skin Imaging Collaboration Project. They also assessed it based on three aspects: keratinocyte carcinoma classification, melanoma classification, and melanoma classification when viewed during dermoscopy. After testing, they saw that the AI’s diagnoses were at par with those made by human dermatologists at 91 percent.
Recently in Seoul, a team of doctors also developed a similar AI. Theirs was trained to learn around 20,000 images of different malignant melanoma and other cancer cells. Their tech produced an accuracy rate of 90 percent in detecting skin cancer.