Smartphone applications that use algorithms to analyse skin lesions may not be very good at determining which ones are cancerous, a new study suggests.
The apps are marketed as educational and so aren't covered as medical devices under the Food and Drug Administration's regulations. But that may not stop some patients from relying on the inexpensive tools, researchers said - which could mean slower diagnosis of potentially dangerous lesions.
"There's no substitute, at this point, for a complete skin exam performed by an expert dermatologist for picking up melanoma as well as other skin cancers," said Dr. Karen Edison, a dermatologist from University of Missouri in Columbia who wasn't involved in the new study.
"Just sending a picture to someone you don't know anywhere in the world can be reassuring if it's very clear that (the lesion) is benign, so that's a good thing," she told Reuters Health, "but it's kind of fraught with other issues that we haven't grappled with adequately, I don't think."
For example, even if an app makes a correct diagnosis of melanoma, that doesn't necessarily help if the patient doesn't know where to get a biopsy or doesn't have insurance to pay for it, Dr Edison said. "We're all for technology, but we need to keep it in perspective, and make it a tool."
For the new study, researchers used photos of 188 pre-diagnosed lesions - 60 melanomas and 128 benign lesions - to check the accuracy of four Smartphone apps made to look for melanoma in previously-taken images.
Three of those apps, which cost under $5 (R40) to own, use algorithms to determine whether a lesion is likely to be cancerous or not. The fourth sends images to a certified dermatologist for evaluation, at a price of $5 per lesion.
Of the three algorithm-based apps, the most accurate still missed 18 of the 60 melanomas, mistakenly classifying them as lower-risk, Dr Laura Ferris from the University of Pittsburgh Medical Center in Pennsylvania and her colleagues reported on Wednesday in JAMA Dermatology.
App users "need to know that that's a pretty big risk to take," Dr Ferris said.
The dermatologist consultation app did better than the others, misdiagnosing just one out of 53 evaluable images of cancerous lesions. All but one of the apps classified more than half of the benign, non-cancerous lesions as problematic. The researchers said they chose not to release the commercial names of the apps evaluated because their purpose was to determine the accuracy of this type of tool, in general.
The website of a similar company that markets an app for skin lesion analysis using real-time photos, SkinVision, cautions customers, "Never disregard professional medical advice, or delay in seeking it, because of something you have read on SkinVision. Do not rely on information from SkinVision instead of seeking professional medical advice."
'Apps can't beat in-person exams'
Likewise, the website for the Mole Detective app says, "Always defer to a medical professional if you feel that a mole looks suspicious. Mole Detective's intent is not to diagnosis but to help you better track the symptoms of melanoma at home."
Dr Ferris said there are certain dermatology apps that can help patients.
"There are apps that will do things like teach you about melanomas," she said. "There are ones that will remind you to do your own skin check - that's great."
Both researchers said teledermatology - giving people who live in rural areas, for example, the chance to consult with a dermatologist through photos or video - can be useful. Dr Edison, for example, once used it to diagnose a farmer living hours away with melanoma during harvest season.
But they agreed that for now - and probably for the foreseeable future - machines and apps can't beat in-person exams when it comes to checking for skin cancer.
(Reuters Health, January 2013)
The 3 types of skin cancer
Risk factors for skin cancer
How to do a skin self-examination