Donate to Science & Enterprise

S&E on Mastodon

S&E on LinkedIn

S&E on Flipboard

Please share Science & Enterprise

Skin Cancer Apps Shown Misleading, Unreliable

Phone photo

(tookapic, Pixabay)

11 Feb. 2020. A review of smartphone apps that photograph and purport to analyze skin conditions for signs of cancer shows the apps are unreliable in detecting cancer. Researchers at University of Birmingham and other institutions in the U.K. describe their review and findings in yesterday’s issue of the journal The BMJ.

The team led by Birmingham bio-statistics professor Jonathan Deeks aims to determine the value of smartphone apps that detect skin cancer by analyzing photographs of moles or other skin lesions taken by a smartphone camera. Skin cancer is a common form of cancer affecting two to three million people worldwide, according to data from 2003 cited by the authors. One form of skin cancer, melanoma, accounts for only four percent of all cases, but results in three-quarters of all skin cancer deaths, because of its likelihood to metastasize, or spread to other parts of the body. All skin cancers, including melanoma, are more treatable if found early, thus early detection and treatment are vital.

Mobile phones offer a promising technology for early detection of skin cancer, because of their widespread use, and built-in camera to capture images of skin abnormalities. Artificial intelligence software is also being developed to analyze images and diagnose abnormalities with machine learning algorithms for signs of skin diseases including cancer. The authors cite a report showing 526 smartphone apps for dermatology emerging by 2017, particularly those with telemedicine as a core feature.

Deeks and colleagues reviewed several international medical indexes for studies evaluating smartphone apps for detecting skin cancer. The researchers focused on assessments of apps taking photos of lesions on the skin and analyzing those images for skin cancer. The team then evaluated the studies with a standard quality assessment tool for research known as QUADAS-2, short for Quality Assessment of Diagnostic Accuracy Studies, version 2, and also looking for reference standards for tissue analysis or expert follow-up.

The researchers review revealed nine studies meeting their criteria, covering six smartphone apps. The studies show more apps (64%) use tissue analysis of lesion images with algorithms, than expert recommendations (36%). As a group, say the authors, the studies show serious failings in their methods, with generally small sample sizes, selective recruitment of participants, and high rates of images that could not be evaluated. Moreover, clinicians, not smartphone users, largely determined whether images of moles or lesions were included in the studies.

The team found two skin cancer apps received European regulatory clearance, known as the CE mark, for Conformit Europenne. One CE-marked app was assessed in two studies, with a combined sample of 252 cases and 61 malignant or pre-malignant melanomas, showing a sensitivity, or return of true positive results, of 80 percent, with 78 percent specificity, or return of true negative results. The second CE-marked app was evaluated by a single study having only 15 participants with five melanoma cases reported, showing low rates of sensitivity, 0 to 20 percent. The other four apps in the study either no longer have working web sites or were taken off the market.

The authors say regulatory bodies need to tighten their criteria for authorizing these kind of apps. “Regulators need to become alert to the potential harm that poorly performing algorithm-based diagnostic or risk monitoring apps create,” says Deeks in a university statement. “We rely on the CE mark as a sign of quality, but the current CE mark assessment processes are not fit for protecting the public against the risks that these apps present.”

The authors note that in the U.S., FDA has a stricter assessment process for medical apps, particularly where users could be harmed if the app does not work as intended. So far, FDA has not cleared any apps assessing skin cancer risk from images.

More from Science & Enterprise:

*     *     *

Comments are closed.