Research Finds Failings in Some Apps Used for the Diagnosis of Skin Cancer
In the scramble to bring successful apps for the diagnosis of skin cancer to market, there is a concern that a lack of testing is risking public safety, according to research led by the University of Birmingham. The research, outlined at the British Association of Dermatologists Annual Meeting, reviewed the medical literature on skin cancer apps to explore the number of apps on the market, ascertain how accurate they are, and what the benefits and limitations of these technologic solutions are.
Examples of the technology used in available apps include teledermatology (which involves sending an image directly to a dermatologist), photo storage (which can be used by individuals to compare photos monthly to look for changes in a mole), and risk calculation (based on color and pattern recognition or on fractal analysis).
Review Findings
The researchers found that some of these apps have a comparatively high success rate for the diagnosis of skin cancer. Teledermatology correctly identified 88% of people with skin cancer and 97% of those with benign lesions. Apps that use fractal theory analysis algorithms (detecting irregularities in a fractal pattern) were the next most successful category—they correctly identified 73% of people with skin cancer and 83% of people with benign lesions. These types of technology have huge potential, as 50% of dermatology referrals in the United Kingdom relate to skin cancer.
However, the researchers point to three major failings with some of the apps: a lack of rigorous published trials to show they work and are safe; a lack of input during the app development from specialists to identify which lesions are suspicious; and flaws in the technology used, namely how the photos are analyzed. The researchers explained that, without specialist input, the apps may not recognize rarer or unusual cancers. Even where the technology is efficient, if it has not been combined with specialist input from a dermatologist, it may not pick up on all symptoms.
An area where color and pattern recognition software apps seem to particularly struggle currently is in recognizing scaly, crusted, ulcerated areas or melanomas that do not produce pigment (amelanotic melanomas). This increases the number of false-negatives and delays treatment.
Some apps that compare images on a monthly basis or advise users to seek dermatologist review, based on a risk calculation, are not able to differentiate between finer details that would be identified using a dermatoscope or in person when touched by a dermatologist. If the app is based on advising patients whether to seek professional advice, they may advise wrongly as they have not correctly identified finer details that may point to cancerous lesion.
There are also criteria that an app cannot always register—for example, if a person aged over 40 develops a newly growing mole. In clinic, this person would be advised that the mole should be removed; however, an app may not be able to provide such personalized advice.
Maria Charalambides, of the University of Birmingham College of Medical and Dental Sciences, who conducted the literature review, said: “Future technology will play a huge part in skin cancer diagnosis. However, until adequate validation and regulation of apps is achieved, members of the public should be cautious when using such apps as they come with risk. Any software that claims to provide a diagnostic element must be subject to rigorous testing and ongoing monitoring.”
The content in this post has not been reviewed by the American Society of Clinical Oncology, Inc. (ASCO®) and does not necessarily reflect the ideas and opinions of ASCO®.
