Feature Story

Using AI for early skin cancer detection

by Melissa Jun Rowley

Using AI for early skin cancer detection

Researchers turn to AI and machine learning to help in early skin cancer detection.

The skin we're in is the only skin we've got. With skin cancer ranking at the top of the most commonly diagnosed type of cancer, detection solutions are highly sought after in the medical field.  Now researchers and scientists are turning to Artificial intelligence (AI) and machine learning to help. This technology has already shown promise in early detection in Alzheimer's and breast cancer.  

Alexander Wong, Professor of Systems Design Engineering at Waterloo in Ontario, Canada, is working with a team using AI to detect melanoma skin cancer through the analysis of skin lesions. While melanoma is less common than some other types of skin cancer, it's the deadliest because it's more likely to grow and spread. But if caught early, Wong says skin cancer is also one of the most treatable forms of cancer.

See also: Using AI to transform medicine

"We're inspired to come up with AI technology that allows you to analyze and help clinicians diagnose cancer at an early stage. This helps the odds of it be readily treated before it is too late.""The other important thing to realize is that skin cancer is one of the more readily analyzable forms of cancer given that it exists on the skin, and not internally," shares Wong. "That's why we're inspired to come up with AI technology that allows you to analyze and help clinicians diagnose cancer at an early stage. This helps the odds of it be readily treated before it is too late."

What can AI see that doctors can't?

As it stands now, dermatologists rely on visual examinations of skin abnormalities, which are subjective to the human eye.

Wong and his team have built an AI-powered imaging system that provides more comprehensive information to doctors, including changes in the concentration and distribution of eumelanin, a chemical that gives skin its color, and hemoglobin, a protein in red blood cells. These are both indicators of melanoma.

Elucid Labs commercialized the system and dermatologists have already begun testing it. Wong says the system should be available to more doctors within the year.

Meanwhile, Stanford University computer scientists claim they've created an AI algorithm that can identify skin cancer as well as a professional doctor. They did this using a database of nearly 130,000 images of moles, rashes and lesions.

Brett Krupel, one of the computer scientists working on the project, explained how the algorithm works. In short, it all comes down to pixels. 

An image is an array of pixels. Each pixel is represented by three numbers: red, green, and blue intensity.  The classifier of the system takes all of these raw numbers as input, and spits out a malignant probability. 

See also: A role for robots in caring for the elderly

Krupel says the classifier has millions of parameters that were tuned from images in the team's training data.

As for whether or not this algorithm be more exact than dermatologists, Krupel said: "We don't claim superiority, but I believe it is possible given a large enough dataset to train on. Open health records would help tremendously."

So what's next?

Wong says AI and machine learning have progressed exponentially within the last few years.

"It [AI} has allowed us to obtain a new level of diagnostic knowledge for aiding clinicians make better decisions beyond our wildest dreams, and I firmly believe that AI will continue to benefit all levels of clinical care," he declares. 


The contents or opinions in this feature are independent and may not necessarily represent the views of Cisco. They are offered in an effort to encourage continuing conversations on a broad range of innovative technology subjects. We welcome your comments and engagement.

We welcome the re-use, republication, and distribution of "The Network" content. Please credit us with the following information: Used with the permission of http://thenetwork.cisco.com/.