Communications of the ACM
AI Beats Doctors at Cancer Diagnoses
Researchers from Germany, France, and the U.S. have demonstrated that a deep learning convolutional neural network can outperform experienced dermatologists at identifying skin cancer.
Credit: Shutterstock.com
Researchers from Germany, France, and the U.S. have demonstrated that a deep learning convolutional neural network (CNN) can outperform experienced dermatologists at detecting skin cancer.
The University of Heidelberg's Holger Haenssle compares the CNN to a child's brain, improving with each training session. To train the network, the team used more than 100,000 images of malignant melanomas and benign skin lesions and moles, with information on the diagnosis for each image.
The CNN missed fewer melanomas and misdiagnosed benign lesions as malignant less often than dermatologists, who accurately detected an average of 86.6% of melanomas, while the CNN correctly identified 95% of melanomas.
Haenssle and his colleagues do not believe the technology will replace dermatologists in skin cancer screening, but suggest it can be used as a supplemental tool, for example, in determining whether to biopsy a lesion.
From U.S. News & World Report
View Full Article
Abstracts Copyright © 2018 Information Inc., Bethesda, Maryland, USA
No entries found