MIT researchers have developed an artificial intelligence device that could help doctors detect melanoma – a type of malignant skin tumor responsible for more than 70 percent of all skin cancer-related deaths worldwide.
Doctors usually need to look at suspected colored lesions (SPLs) to determine if they are dangerous and indicative of skin cancer. It is often challenging and due to the many pigmented lesions that need to be examined.
If melanoma is diagnosed early, not only is it easier to treat, but it can also reduce the cost of treatment.
The researchers artificially created a deep convoluted neural network (DCNN). They applied it to analyze the SPLs using extensive field photography, as we took it from our smartphones and personal cameras.
The smartphone camera’s large area image shows a clear vision of a large area of skin. DCNNs then analyzed pigmented skin lesions to identify and screen for early-stage melanoma.
The system examines and marks colored lesions (yellow = consider a further investigation, red = further examination or referral to a dermatologist).
The extracted features use to evaluate other colored lesions and display the results as a heat map.
Researchers at the Gregorio Marañón Hospital in Madrid have studied the system using 20,388 wide-field images of 133 patients with publicly available photos. Dermatologists work with researchers to visualize injuries in pictures for comparison.
They found that the system achieved more than 90.3 percent accuracy in separating the SPLs from unnecessary wounds.
“Valuable research on SPLs can save lives; However, the current medical system lacks the capacity to provide comprehensive skin tests on a large scale,” said Louis R. Soenksen, a doctoral and medical device specialist.
“Our research shows that systems that use computer vision and deep neural networks to measure these common signals can achieve comparable accuracy to that of a dermatologist,” explains Soenksen.
“We hope our study revives the desire to provide more efficient dermatological examinations in primary care settings to encourage referrals.”
Doctors or patients can also use screening.