AI diagnoses skin cancer more accurately than dermatologists
A new branch of artificial intelligence (AI), known as a deep learning convolutional neural network (CNN), has proven to better detect and diagnose skin cancer than experienced dermatologists.
When compared to the performance of 58 dermatologists from 17 different countries around the world, roughly half of whom only had more than five years of experience, CNN missed fewer melanomas and misdiagnosed benign moles less often, according to a study published May 29 in Annals of Oncology.
For their study, a team of international researchers trained a CNN by showing it more than 100,000 dermoscopic images of malignant melanomas and benign skin moles.
“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,” explained lead author Holger Haenssle, MD, PhD, from Heidelberg University in Germany, in a prepared statement.
The group of international dermatologists were asked to perform two tasks: make a diagnosis and management decision of malignant melanoma or benign mole just from the dermoscopic images (level one) and then four weeks later after being given clinical information about the patient and close-up images of the same 100 cases (level two), make diagnoses and management decisions again.
For level one, dermatologists accurately detected an average of 86.6 percent of melanomas and an average of 71.3 percent of benign lesions. However, the CNN outperformed the dermatologists by detecting 95 percent of melanomas.
For level two, the dermatologists improved by diagnosing 88.9 percent of malignant melanomas and 75.7 percent benign when given additional clinical information and close-up images, but again were outperformed by the CNN.
“These findings show that deep learning convolutional neural networks (CNNs) are capable of out-performing dermatologists, including extensively trained experts, in the task of detecting melanomas,” Haenssle said.
Haenssle and his colleagues noted that CNN could be used as an additional aid to dermatologists and not replace them entirely.
“This CNN may serve physicians involved in skin cancer screening as an aid in their decision whether to biopsy a lesion or not," Haenssle said. "Most dermatologists already use digital dermoscopy systems to image and store lesions for documentation and follow-up. The CNN can then easily and rapidly evaluate the stored image for an ‘expert opinion’ on the probability of melanoma. We are currently planning prospective studies to assess the real-life impact of the CNN for physicians and patients.”
A recent graduate from Dominican University (IL) with a bachelor’s in journalism, Melissa joined TriMed’s Chicago team in 2017 covering all aspects of health imaging. She’s a fan of singing and playing guitar, elephants, a good cup of tea, and her golden retriever Cooper.