Trial of novel image-enhancement software to assist medical students in identifying dermoscopic features of melanoma.

BACKGROUNDEarly detection of melanoma is central to reducing the disease-related burden. The dermatoscope is used by skin cancer clinicians to help identify features of melanoma in a skin lesion. Features can be subtle and identification can be a challenge for early dermoscopists.

OBJECTIVESTo investigate the utility of novel image-enhancement software as a possible aid for beginner dermoscopists in identifying melanoma features. The software had previously been developed for optometry for the diagnosis and monitoring of anterior eye conditions.

METHODSFollowing an introductory lecture on dermoscopy, third-year medical students at James Cook University were asked to complete a 10-lesion test set of dermoscopic images on benign lesions and cutaneous melanoma (CM). Students scored the presence or absence of some melanoma-related features (asymmetry, blue-grey colour, atypical network) and made a diagnosis for a standard image and then for the matching enhanced image.

RESULTSIn total, 61 students participated. Total performance scores did not differ significantly between standard and enhanced images of the benign lesions. However, students correctly identified CM-related features (blue-grey colour P = 0.01, atypical network P < 0.001) and diagnosed melanoma correctly (P = 0.005) significantly more often when observing the enhanced images of CM compared with standard images. Participants also reported a subjective improvement in diagnostic confidence when using the enhanced images.

CONCLUSIONSThe image-enhancement software shows promise in assisting beginner dermoscopists in identifying some features of melanoma and could be tested with more advanced dermoscopists along with a wider range of dermoscopic features. The software could potentially be made widely and cheaply available in the form of a mobile phone application or be incorporated into an existing image-management platform.

Abstract from PubMed