The content of this website is intended for healthcare professionals only

A.I. better at diagnosing skin cancer than humans

Machines detected 95% of melanomas while humans spotted 86.6%

Adrian O'Dowd

Tuesday, 29 May 2018

An artificial intelligence (AI) system appears to be better at spotting and diagnosing skin cancer than human physicians, claims a study* published today in the Annals of Oncology.

A team of international researchers have now theorised that such technology could become an additional aid for dermatologists in diagnosing skin cancers.

Malignant melanoma incidence is increasing, with an estimated 232,000 new cases worldwide and around 55,500 deaths from the disease each year even though it can be cured if detected early.

A form of AI, or machine learning, known as a deep learning convolutional neural network (CNN) can help with melanoma detection, but data comparing a CNN’s diagnostic performance to larger groups of dermatologists are lacking.

Therefore researchers in Germany, the USA and France trained a CNN to identify skin cancer by showing it more than 100,000 images of malignant melanomas (the most lethal form of skin cancer), as well as benign moles (or nevi).

They created two test sets of images from the image library of the Department of Dermatology at University of Heidelberg in Germany 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 and before doing so, 100 of the most difficult lesions were selected to test real dermatologists in comparison to the results of the CNN.

For the study, dermatologists from around the world were invited to take part, and 58 from 17 countries around the world agreed.

Of these, 29% indicated they had less than two years’ experience in dermoscopy, 19% said they were skilled with between two to five years’ experience, and 52% were expert with more than five years’ experience.

The dermatologists were asked to first make a diagnosis of 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 (including age, sex and position of the lesion) and close-up images of the same 100 cases (level II) and asked for diagnoses and management decisions again.

In level I, the dermatologists accurately detected an average of 86.6% of melanomas, and correctly identified an average of 71.3% of lesions that were not malignant.

However, when the CNN was tuned to the same level as the physicians to correctly identify benign moles (71.3%), the CNN detected 95% of melanomas. At level II, the dermatologists improved their performance, accurately diagnosing 88.9% of malignant melanomas and 75.7% that were not cancer.

The first author of the study, Professor Holger Haenssle, senior managing physician at the Department of Dermatology, University of Heidelberg, 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.”

The researchers concluded: “In conjunction with results from the reader study level-I and -II we could show, that the CNN’s diagnostic performance was superior to most but not all dermatologists.

“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. Therefore, physicians of all different levels of training and experience may benefit from assistance by a CNN’s image classification.”

*H.A. Haenssle, et al. Man against machine: diagnostic performance of a deep learning convolutional neural network for dermoscopic melanoma recognition in comparison to 58 dermatologists. Annals of Oncology. DOI:10.1093/annonc/mdy166

Registered in England and Wales. Reg No. 2530185. c/o Wilmington plc, 5th Floor, 10 Whitechapel High Street, London E1 8QS. Reg No. 30158470