Imaging Software Detects Cancer With Human Accuracy
Research by Stanford University could be a game changer in skin cancer detection.
The medical industry faces a horrible conundrum. It’s not just the ever-increasing costs, obstacles preventing access to quality care, and lack of patient awareness, either. Too often, issues that could have been treated with a better success rate and lower cost instead escalate towards mortality – and near bankruptcy – if not addressed early enough.
One solution has been the technological innovation that makes early detection more affordable and less time consuming. Think back to the invention of the “at home” pregnancy test, for example. It eliminated the need to make a doctor’s appointment to confirm, and prior to its widespread marketing, women were counseled not to even make that appointment until they were as much as three months along. Now, with inexpensive tests that can detect pregnancies that are a matter of days along, women are better able to take charge of their prenatal care and can expect healthier outcomes.
One exciting new innovation by researchers at Stanford University stands to be a game changer in skin cancer detection. This software, which can literally be loaded onto a smartphone and use the phone’s camera, is able to identify skin cancers with approximately the same accuracy as the human doctor operating it. While both the doctor and the software would still need to rely on a biopsy for a definitive diagnosis, this new tool could potentially put the power of pre-screening into more hands, as well as do wonders for those cases of “let’s just watch this mole and report any changes.”
According to a report from Stanford News, “We realized it was feasible, not just to do something well, but as well as a human dermatologist,” said Sebastian Thrun, an adjunct professor in the Stanford Artificial Intelligence Laboratory. “That’s when our thinking changed. That’s when we said, ‘Look, this is not just a class project for students, this is an opportunity to do something great for humanity.”
In a similar methodology as facial recognition or voice recognition software being “trained” through the use of a plethora of samples, the software was fed images of as many as 130,000 different skin samples. In this way, it learned to select lesions that look suspicious and therefore indicate the need for further testing.