Results of the 2016 International Skin Imaging Collaboration International Symposium on Biomedical Imaging challenge: Comparison of the accuracy of computer algorithms to dermatologists for the diagnosis of melanoma from dermoscopic images

Abstract

Deep learning computer vision systems classified melanoma dermoscopy images with accuracy that exceeded some but not all dermatologists.

Deep‐learning‐based, computer‐aided classifier developed with a small dataset of clinical images surpasses board‐certified dermatologists in skin tumour diagnosis

“…However, the use of clinical images (usually taken with digital cameras) as the input to train artificial intelligence for CADx has been considered problematic because digital images carry limited morphological information . To achieve high sensitivity and specificity in such systems, most previous studies used images from dermoscopy, multispectral imaging systems, reflectance confocal microscopy and optical coherence tomography . Several systems have already been introduced to the market and have reported high sensitivity .…”

Validation of an integrated dermoscopic scoring method in an European teledermoscopy web platform: the iDScore project for early detection of melanoma

“…[20][21][22][23][24][25][26][27] Although these tools are increasingly developed, paradoxically they are poorly applied in clinical practice, probably due to the difficulty of data interpretation and physician distrust. 26,27 The key point is that dermatologists are still superior to computers because unlike current CAD models, they simultaneously evaluate medical history and the clinical data of suspected MSLs. 26,27 Scoring classifier algorithms that combine relevant clinical, historical and dermoscopic data have been demonstrated to be a valid alternative to CAD systems.…”

“…26,27 The key point is that dermatologists are still superior to computers because unlike current CAD models, they simultaneously evaluate medical history and the clinical data of suspected MSLs. 26,27 Scoring classifier algorithms that combine relevant clinical, historical and dermoscopic data have been demonstrated to be a valid alternative to CAD systems. 13,[28][29][30] We can therefore confirm the iDScore as a rapid and easy tool that helps dermatologists of any skill level to manage suspected MSLs in routine clinical practice by rapidly calculating the lesion score and consigning a high-confidence noninvasive differential diagnosis between AN and EM, whether face to face or in a teledermoscopy setting.…”

How can emergency physicians harness the power of new technologies in clinical practice and education?

“…This brings a reality of diagnostics and therapeutics being reduced to a computer program of questions, an input of laboratory results and a computerised analysis of radiological investigations. Such technologies are already in late development, for example, computer algorithms may be able to match dermatologists' assessment of melanomas35 using smartphone technologies. Could such a march of technology lead to the tech-Emergency Department (ER) of the future with electronic diagnostics and intubation performing robots?…”