An international team of researchers say they have developed an artificial intelligence system that diagnoses skin cancer more accurately than trained dermatologists.
For the study, a team of researchers from Germany, the U.S. and France trained an AI system to identify skin cancer using 100,000-plus images, including images of malignant melanomas and benign moles. The researchers used a type of AI known as a convolutional neural network, which learns over time, rather than having to be programmed like typical software.
The researchers presented the trained convolutional neural network with 300 new images to determine whether the AI system was able to differentiate between cancerous and noncancerous skin. The researchers found the convolutional neural network successfully identified 95 percent of the melanomas and 71.3 percent of the benign moles, according to results published in the journal Annals of Oncology.
To assess the outcomes of the convolutional neural network, the researchers also enrolled 58 dermatologists from 17 countries to evaluate 100 of the cases. When provided with clinical information and close-up images of the cases, the dermatologists were able to accurately diagnose 88.9 percent of malignant melanomas and 75.7 percent of benign moles.
The researchers concluded the convolutional neural network missed fewer melanomas and misdiagnosed benign moles less frequently than the dermatologists. This study marks the first time a form of AI has reported better results at detecting melanomas than trained dermatologists, according to a May 28 statement from the European Society for Medical Oncology.
“These findings show that deep learning convolutional neural networks are capable of out-performing dermatologists, including extensively trained experts, in the task of detecting melanomas,” Holger Haenssle, a senior managing physician in the dermatology department at University of Heidelberg in Germany and first author of the study, said in the statement.
For more information: CLICKHERE