A collaborative study by researchers in Germany, France and the US has shown that a form of artificial intelligence (AI) known as a deep learning convolutional neural network (CNN) can detect skin cancer better than experienced dermatologists.
The study has been published in Annals of Oncology and explains how the researchers trained the CNN. Its performance was compared to that of 58 international dermatologists. CNN missed fewer melanomas and misdiagnosed benign moles as malignant less often than the human clinicians.
The first author of the study is Professor Holger Haenssle, senior managing physician at the Department of Dermatology at the University of Heidelberg in Germany. He said: “The CNN works like the brain of a child. To train it, we showed the CNN more than 100,000 images of malignant and benign skin cancers and moles and indicated the diagnosis for each image.
“Only dermoscopic images were used, that is lesions that were imaged at a 10-fold magnification. With each training image, the CNN improved its ability to differentiate between benign and malignant lesions.
“After finishing the training, we created two test sets of images from the Heidelberg library that had never been used for training and therefore were unknown to the CNN. One set of 300 images was built to solely test the performance of the CNN. Before doing so, 100 of the most difficult lesions were selected to test real dermatologists in comparison to the results of the CNN.”
The participating dermatologists were first asked to diagnose malignant melanoma or benign mole just from the dermoscopic images (level I) and make a decision about how to manage the condition. Four weeks later they were given clinical information about the patient and close-up images of the same 100 cases (level II) and asked to revaluate their diagnoses and management decisions.
In level I, the dermatologists accurately detected an average of 86.6% of melanomas, and correctly identified around 71.3% of lesions that were not malignant. However, when the CNN was tuned to the same level it detected 95% of melanomas. At level II, the dermatologists improved their performance, accurately diagnosing 88.9% of malignant melanomas and 75.7% of benign moles.
Haenssle said: “The CNN missed fewer melanomas, meaning it had a higher sensitivity than the dermatologists, and it misdiagnosed fewer benign moles as malignant melanoma, which means it had a higher specificity; this would result in less unnecessary surgery.
“When dermatologists received more clinical information and images at level II, their diagnostic performance improved. However, the CNN, which was still working solely from the dermoscopic images with no additional clinical information, continued to out-perform the physicians’ diagnostic abilities.”
The study did have several limitations. These included the fact that the dermatologists were in an artificial setting and knew they were not making life-or-death decisions, the test sets did not include a full range of skin lesions and there were fewer validated images from non-Caucasian skin types and genetic backgrounds.
The researchers do not imagine CNN will take over from dermatologists in diagnosing skin cancers, but think it could be used as an additional aid. They are currently planning studies to evaluate the real-life impact of the CNN for physicians and patients.
In the UK, Prime Minister Theresa May recently announced plans to increase the use of data and AI to transform the diagnosis of chronic diseases, with the aim of 22,000 fewer people dying from cancer each year by 2033.