Could Smartphones be Used to Detect Skin Cancer?
Possibly. A recent study in the journal Nature found that computers programmed with image recognition technology could diagnose skin cancer just as accurately as skilled skin cancer doctors evaluating the same images.
These doctors, known as dermatologists, are trained to identify skin cancer, including melanoma, the deadliest form of skin cancer. They often use the ABCDE system to assess features on the skin to distinguish between a regular mole and melanoma.
Stanford University researchers “trained” a computer to diagnose skin cancer by classifying digital images of skin lesions. They provided the computer with 129,450 images of skin lesions that were labeled with the name of the skin disease that each image represented. These images covered 2,032 diagnosed skin diseases.
The researchers taught the computer by using a technique called “deep learning.” The more images provided to the computer, the better it gets at making a diagnosis.
The researchers then tested how precise the computer was at identifying skin cancer. They gave the computer previously unseen skin lesion images and asked whether the lesions needed further medical attention. In parallel, the researchers showed the same set of images to 21 skilled dermatologists.
The computer did as well or better than all the doctors in diagnosing the provided images and accurately distinguishing between cancerous and non-cancerous skin lesions.
However, in fairness to the doctors, the researchers did not test whether the doctors are more effective at diagnosing skin disease using traditional physical skin exams instead of only digital images.
This study isn’t the only example of artificial intelligence being used in making medical decisions for cancer. The IBM super computer Watson, best known for beating Jeopardy champion Ken Jennings, is being used at a Florida hospital to interpret clinical information of cancer patients to identify the best treatment options.
The possibilities of this current study are profound. It offers an effective, cheap and easy way for doctors and patients to track skin lesions and detect cancer earlier. This is extremely important, since the survival rate for melanoma is less than 20 per cent when it spreads to other organs or distant skin areas.
However, before you start searching for the app, more research needs to be done before it can be made more widely available. For instance, we need to know if the computer can distinguish between similar looking skin diseases. Would it be helpful with decision making in the clinic? How would it fare when faced with the full spectrum of skin lesions seen in the clinic? The potential for this research and its implications on the future of early cancer detection is tremendous.
Kelly Fathers, PhD
Read more: http://www.cancer.ca/en/research-horizons/2/8/1/could-smartphones-be-used-to-detect-skin-cancer/?region=on#ixzz4dyaFvcJt