Why cancer-identifying apps aren't ready for prime time
Of four popular smartphone applications used to ID melanoma, only one passed the test
The premise is promising, at least on paper: You use your smartphone to quickly snap a photo of a questionable mole or skin lesion. Next, an algorithm or anonymous, board-certified dermatologist scans it for tell-tale signs of melanoma — things like size, asymmetry, an irregular border, or odd color. The intent is to assuage any fears a person might have about that strange thing they have growing on them, potentially steering them towards the correct course of treatment, if necessary.
But a new study that analyzes popular cancer-identifying apps says that only one is reliably accurate in most cases. Researchers at the University of Pittsburgh School of Medicine used photos of 188 moles, some confirmed with melanoma, to test four apps. Three are free (or close to free) and use algorithms. The fourth, which costs $5 per photo and has a 24-hour turnaround, beams the photo to a dermatologist.
The dermatologist app, SkinVision, correctly identified suspicious moles 98 percent of the time. The other, algorithm-based applications missed melanomas 30 percent of the time, which is worrisome for a treatable skin disease that relies on early detection.
These pre-screening tools could be giving people a "false sense of security," study co-author Laura Ferris told NPR. While these apps have good intentions, at this point the technology is "not good enough to trust your life with." (Via CBS, NPR)