Accuracy of skin cancer diagnostic apps


  • Heather Mason
  • Univadis Medical News
L'accesso ai contenuti di questo sito è riservato agli operatori del settore sanitario italiano L'accesso ai contenuti di questo sito è riservato agli operatori del settore sanitario italiano

Skin cancer smartphone applications (apps) provide a technological approach to assist people with suspicious lesions whether to seek further medical opinion. Some apps forward images to an experienced professional for review, while others have in-built algorithms (artificial intelligence) that classify images into high or low risk. Conformit Europenne (CE) marking has been applied, allowing distribution of two algorithm-based apps in Europe.

A systematic review published in the British Medical Journal identified nine studies; six studies (725 skin lesions) evaluated diagnostic accuracy by comparing with histopathological reference standard diagnosis. Five of these detected melanoma only and one aimed to distinguish between malignant and benign lesions. A further three studies (407 lesions) verified the app against expert recommendations.

SkinScan (n=15, five melanomas) had 0 per cent sensitivity but 100 per cent specificity for the detection of melanoma. SkinVision (n=252, 61 malignant/premalignant lesions) had sensitivity of 80 per cent (95% CI 63%-92%) and specificity of 78 per cent (95% CI 67%-87%) for the detection of malignant/premalignant lesions. The accuracy of the SkinVision app was poor when verified against expert recommendations (three studies).

The authors concluded that current apps are unreliable in detecting melanoma or other skin cancers. Awarding a CE mark does not provide adequate protection to the public, they said.