AI outperforms dermatologists in diagnosing skin cancer
A convolutional neural network was trained using more than 100,000 images of malignant melanomas and benign moles.
Dermatologists are no match for artificial intelligence when it comes to diagnosing skin cancer, according to a new study by researchers in the United States, France and Germany.
The international team trained a convolutional neural network (CNN) to identify skin cancer by showing it more than 100,000 images of malignant melanomas as well as benign moles.
Specifically, they trained and validated Google’s Inception v4 CNN architecture using dermoscopic images at a 10-fold magnification and corresponding diagnoses. Then, they compared its performance with that of 58 international dermatologists from 17 countries—including 30 experts with more than five years of experience.
Results of the study, published this week in the Annals of Oncology, show that the CNN missed fewer melanomas and misdiagnosed benign moles as malignant less often than the group of experienced dermatologists.
“These findings show that deep learning convolutional neural networks are capable of outperforming dermatologists, including extensively trained experts, in the task of detecting melanomas,” said Holger Haenssle, first author of the study and senior managing physician at the University of Heidelberg’s Department of Dermatology.
Also See: Melanoma app vendor barred from making deceptive health claims
The authors conclude the results of their study “demonstrate that an adequately trained deep learning CNN is capable of a highly accurate diagnostic classification of dermoscopic images of melanocytic origin” and that “physicians of all different levels of training and experience may benefit from assistance by a CNN’s image classification.”
They also observe that “while a CNN’s architecture is difficult to set up and train, its implementation on digital dermoscopy systems or smart phone applications may easily be deployed.”
Aadi Kalloo, an author of the study and senior data analyst at New York’s Memorial Sloan Kettering Cancer Center, contends that “a human definitely always needs to be in the loop” when diagnosing skin cancer, adding that the CNN’s diagnostic performance was superior to most, but not all, dermatologists. However, he says that the study’s results are a “good starting point for computer-aided diagnosis,” which will ultimately “help speed things up in the clinic and bring costs down in the long run.”
Kalloo notes that the dataset of images used for CNN testing in comparison to the 58 dermatologists came from the International Skin Imaging Collaboration (ISIC) archive, which contains the largest publicly available collection of quality-controlled dermoscopic images of skin lesions.
The ISIC archive, led by Sloan Kettering, currently contains more than 34,000 dermoscopic images collected from leading dermatology centers around the world, according to Kalloo, who is the primary data manager. “We have plans to upload tens of thousands more (images) this year,” he concludes.
The international team trained a convolutional neural network (CNN) to identify skin cancer by showing it more than 100,000 images of malignant melanomas as well as benign moles.
Specifically, they trained and validated Google’s Inception v4 CNN architecture using dermoscopic images at a 10-fold magnification and corresponding diagnoses. Then, they compared its performance with that of 58 international dermatologists from 17 countries—including 30 experts with more than five years of experience.
Results of the study, published this week in the Annals of Oncology, show that the CNN missed fewer melanomas and misdiagnosed benign moles as malignant less often than the group of experienced dermatologists.
“These findings show that deep learning convolutional neural networks are capable of outperforming dermatologists, including extensively trained experts, in the task of detecting melanomas,” said Holger Haenssle, first author of the study and senior managing physician at the University of Heidelberg’s Department of Dermatology.
Also See: Melanoma app vendor barred from making deceptive health claims
The authors conclude the results of their study “demonstrate that an adequately trained deep learning CNN is capable of a highly accurate diagnostic classification of dermoscopic images of melanocytic origin” and that “physicians of all different levels of training and experience may benefit from assistance by a CNN’s image classification.”
They also observe that “while a CNN’s architecture is difficult to set up and train, its implementation on digital dermoscopy systems or smart phone applications may easily be deployed.”
Aadi Kalloo, an author of the study and senior data analyst at New York’s Memorial Sloan Kettering Cancer Center, contends that “a human definitely always needs to be in the loop” when diagnosing skin cancer, adding that the CNN’s diagnostic performance was superior to most, but not all, dermatologists. However, he says that the study’s results are a “good starting point for computer-aided diagnosis,” which will ultimately “help speed things up in the clinic and bring costs down in the long run.”Kalloo notes that the dataset of images used for CNN testing in comparison to the 58 dermatologists came from the International Skin Imaging Collaboration (ISIC) archive, which contains the largest publicly available collection of quality-controlled dermoscopic images of skin lesions.
The ISIC archive, led by Sloan Kettering, currently contains more than 34,000 dermoscopic images collected from leading dermatology centers around the world, according to Kalloo, who is the primary data manager. “We have plans to upload tens of thousands more (images) this year,” he concludes.
More for you
Loading data for hdm_tax_topic #better-outcomes...