Skin cancer diagnosis apps 'unreliable', missing deadly skin cancers - study

Smartphone apps dubbed as 'early warning systems' for detecting skin cancer are poorly regulated and shouldn't be trusted, according to a new study out of the UK. 

Gaining popularity in the last few years, a handful of smartphone apps and devices claim to analyse moles and skin lesions through the use of artificial intelligence.

But researchers at the University of Nottingham have dubbed them unreliable and even dangerous in a new study published this week in the British Medical Journal. 

"These apps are providing information that could lead to potentially life-or-death decisions," says co-lead author Dr Hywel C Williams. 

SkinVision and SkinScan are two of the most readily available and popular mole-analysing apps. 

Researchers evaluated the apps by seeing how effectively they picked dangerous melanomas out of groups of moles. 

SkinScan was evaluated in a study of 15 moles with five melanomas. It failed to identify any of the melanomas.

SkinVision was evaluated in two studies. Out of 108 moles, it only recognised 88 percent of the dangerous or pre-cancerous ones, while 21 percent of the non-problematic moles were wrongly identified as being cancerous. 

"Algorithm-based smartphone apps cannot be relied on to detect all cases of melanoma or other skin cancers," the study concluded. 

Researchers also pointed out that their performance when being used by the public was also probably likely to be poorer than when done in a clinical setting, meaning they don't "provide adequate protection to the public". 

Previously, the American Federal Trade Commission fined the marketers of two other apps - MelApp and Mole Detective - for "deceptively claiming the apps accurately analysed melanoma risk".