When it comes to melanoma, an early diagnosis often means the difference between life and death. Researchers at Stanford University have now developed a computer program that can detect skin cancer with as much accuracy as a dermatologist. Their goal was to increase access to important skin cancer screening.
The researchers used deep learning and visual processing to develop the program, which references about 130,000 images of skin lesions. First, they created a comprehensive database of images using the internet and information from Stanford's medical school. Then, they selected specific high-quality images representing about 2,000 skin conditions. Finally, the researchers developed the algorithm and tested the program using a scientific curve that differentiates benign lesions from those that are cancerous.
Testing showed that the program's success matched the performance of the 21 dermatologists who participated in the study. In the future, researchers hope to develop a smartphone app using the algorithm, making it accessible to more people.
Image Credit: Thinkstock