Using Artificial Intelligence to Diagnose Skin Cancer Might Not Be So Far In the Futureby Eileen Bailey Health Writer
The possibility of skin cancer is scary. But what if your doctor could channel information from doctors all over the world to find out more about your cancer, whether you need a biopsy, if you have cancer, what stage it is in, and what is the best treatment. Imagine your doctor could search through thousands, or hundreds of thousands, of images of skin cancer, and discover what would work best with your particular case.
That is exactly what researchers are trying to accomplish by incorporating artificial intelligence (AI) into the diagnosis and treatment of skin cancer.
Artificial intelligence in medicine
Computers are frequently used to analyze a set of data. Scientists are taking that one step further and teaching computers “deep learning” by building algorithms so that computers can think for themselves and become better at analyzing a set of data. How could that work in medicine? In skin cancer, using hundreds of thousands of images and applying algorithms, a program could provide information on diagnosis and stage of a single lesion within minutes. Deep learning provides a way for that program to continuously improve its analysis of images. The computer can learn about biological changes that can’t be seen or understood by looking with just the human eyes. The computer might notice tiny biological changes that could signal a potential for disease.
This research is taking place in another field of medicine: cardiology. Forbes Magazine describes a study completed in the UK where health information for 82,000 patients was analyzed using computer algorithms to predict future heart attacks. The results were compared to predictions based on criteria from the American College of Cardiology/American Heart Association, such as age, smoking history, history of diabetes, cholesterol levels, and other criteria. The computers did a better job at predicting which people would have a heart attack than humans.
Scientists are now working on using this type of technology for recognizing skin cancer, tuberculosis, breast cancer, and in the fields of pathology, ophthalmology, and cardiology, according to Forbes Magazine. AI could revolutionize health care by diagnosing diseases earlier and identifying people at risk before a disease develops.
AI for skin cancer
AI technology for skin cancer could be available as early as 2019. At the University of Waterloo in Canada, scientists are developing technology to help detect melanoma during the early stages. Melanoma is the deadliest form of skin cancer; however, the five-year survival rate is 97 percent when detected early and 14 percent when detected in late stages, according to the American Cancer Society. This makes early detection vital. The researchers at the University of Waterloo have entered tens of thousands of images of melanoma lesions, along with corresponding eumelanin and hemoglobin levels into a computer program. The program then analyzes the characteristics and biomarkers of a lesion to determine whether it could be melanoma. This has the potential to decrease the number of biopsies needed, as well as reduce the waiting time to find out a diagnosis.
Researchers at Stanford University are also working on AI to detect skin cancer. They created a database of about 130,000 skin cancer images and created an algorithm for a computer to compare a lesion with these images and determine the probability of cancer. When tested, this program matched the performance of dermatologists. Their program used an algorithm developed by Google that identified images and placed them into one of 1,000 categories. They gathered images of skin cancer lesions from the internet to create a large database. The computer looks at the new image, compares it to the images in the database and puts it into a category of over 2,000 different skin diseases. They hope their research can be used to “visually screen for skin cancer and other diseases.”
Apps on your phone
There are several apps available for smart phones to help people track moles and other lesions. Some of these apps can alert you that you should see a dermatologist. At first, the AI programs that are being developed are meant to be used by doctors as an aid in identification and diagnosis of skin cancer, but in the future they might be available on your phone. There is some concern about the ability of apps on your smart phone providing a diagnosis or even a potential diagnosis. Will people use this information to avoid seeing a dermatologist or seeking treatment? Will people who receive a false negative reading not seek treatment when they should? Or will these types of programs help by increasing awareness and alerting people to a potential problem that they might have ignored in the past?
One such app, SkinVision, is not yet available in the United States, but according to the company website, is “coming to the U.S. soon.” This app instantly analyzes a spot and gives you a recommendation. Other skin cancer apps, such as the University of Michigan’s UMSkinCheck, offer a way to store images of a full body scan and spots or moles to monitor changes. It also reminds you when it is time to complete a new scan. It’s meant to raise awareness, provide a place to track changes, and improve communication with your dermatologist.