Scientists at Stanford University, USA, have developed a method for diagnosing skin cancer with artificial intelligence (AI) that could also be used in smartphones.
Using an artificially intelligent diagnosis algorithm for skin cancer, the team of scientists created a database of roughly 130,000 skin disease images and trained their algorithm to visually diagnose potential cancer. Following a series of tests the team saw that the AI performed with “inspiring accuracy”, which led to a final test against 21 board-certified dermatologists. "In its diagnoses of skin lesions, which represented the most common and deadliest skin cancers, the algorithm matched the performance of dermatologists,” the team said. Its success, which is documented in the journal Nature, has encouraged the team to make this technology available for smartphones.
Tackling skin cancer
The team chose skin cancer because every year there are roughly 5.4 million new cases of skin cancer in the United States. Survival rates are around 97 percent if the melanoma is detected early, however. The AI was developed with an existing algorithm developed by Google that was already trained to differentiate images such as cats and dogs. To teach the machine to identify skin cancer, it was fed 130,000 images of skin lesions representing over 2,000 different diseases. This enabled it to identify the most common and deadliest skin cancers, malignant carcinomas and malignant melanomas. In three tests against dermatologists, the AI algorithm matched the performance of its human counterparts.
AI by smartphone
The algorithm currently exists on a computer and cannot currently make a full diagnosis, as this is normally confirmed with a tissue biopsy, but the team has plans to take it further.
One connected device it could be used on is the smartphone.
“My main eureka moment was when I realized just how ubiquitous smartphones will be,” said Dr Andre Esteva, co-lead author of the paper. “Everyone will have a supercomputer in their pockets with a number of sensors in it, including a camera. What if we could use it to visually screen for skin cancer? Or other ailments?” Until then, the algorithm will be used to assist human doctors in their diagnoses. “Advances in computer-aided classification of benign versus malignant skin lesions could greatly assist dermatologists in improved diagnosis for challenging lesions and provide better management options for patients,” said Susan Swetter, professor of dermatology and director of the Pigmented Lesion and Melanoma Program at the Stanford Cancer Institute, and co-author of the paper. “However, rigorous prospective validation of the algorithm is necessary before it can be implemented in clinical practice, by practitioners and patients alike.”
What does this mean for healthcare professionals?
With AI once again proving its worth in the medical world, how far away are we from AIs replacing doctors?
IBM Watson has already shown success in diagnosing rare diseases, while the UK’s NHS is trialing an AI chat-bot to help improve triage for non-life-threatening conditions. Digital health futurist, Maneesh Juneja, told Internet of Business that, though interesting, this does not change how we diagnose skin cancer in the immediate future.
“There will need to be validation of these algorithms before anyone in healthcare can use them, and clinical validation takes time. My takeaway from this research is that the pace of change is accelerating, and it’s important to have an open mindset towards what machines might be able to do better than us over the long term.”
Striking a similar tone, David Doherty – cofounder of 3G Doctor – told IoB that this is “another great example of the potential we have in the healthcare industry to use tools that are already being used at scale in the mobile industry.”
“Dermatologists should be clamoring to use software like this even if just as a safety check process but I’m not optimistic they will as most don’t yet even accept photos from patients and yet we know more than 3 billion citizens are carrying a camera phone!” he said. When asked whether AIs might replace doctors soon, Robert van der Veur, project manager IoT, BI & Analytics at Sogeti, replied: “A doctor does much more than only diagnosing patients, like curing them for example. This type of AI is only a tool, similar to a cat scan machine, that supports doctors being better in their job.” It seems that, for now, diagnoses made by AI will “ultimately still require clinical judgement”, as Loy Lobo, founder of Wegyanik, told us late last year.