AI Skin Cancer Checks: DermaSensor, DERM and Mole-Checking Apps Compared

Three very different kinds of AI now look at moles and skin spots. They are not equally proven, and only one of them is meant for you to use alone.

Why this matters

Skin cancer is one of the most common cancers, and most skin cancers are curable when found early. The problem is access: in England alone, more than 650,000 people were referred on the urgent suspected skin cancer pathway in 2023/24, and 120,000 of them did not have cancer confirmed or ruled out within the 28-day standard, according to NHS England. Most referred spots turn out to be harmless. That makes skin checking an obvious target for AI: if software can safely clear benign lesions, specialists can spend their time on the patients who need them.

But "AI skin check" covers very different products, with different users, regulators and evidence. This guide separates them.

Three kinds of AI skin tools

DermaSensor: a second opinion for family doctors

DermaSensor is a handheld device that shines light into a skin lesion and measures how it scatters (a technique called elastic scattering spectroscopy). An algorithm turns the readings into a result: investigate further, or monitor. The FDA granted it De Novo authorisation on 12 January 2024 creating a new device type for AI-assisted skin cancer assessment in primary care.

What it is cleared for

According to the FDA decision summary, it is for doctors who are not dermatologists, for patients aged 40 and over, and only for lesions that already look suspicious for melanoma, basal cell carcinoma or squamous cell carcinoma. It is an adjunct to help decide whether to refer, not a stand-alone diagnosis and not a screening tool for checking every mole.

What the evidence shows

In the DERM-SUCCESS pivotal study (1,005 participants, 1,579 lesions, 22 sites), the device detected 95.5% of skin cancers, compared with 83.0% for the primary care doctors. But its specificity was 20.7%, against 54.2% for the doctors: roughly four in five benign lesions were still flagged. In reader studies, doctors using the device caught more cancers but also referred more harmless spots.

In practice, that profile suits a cautious "rule-in for referral" tool. It is not designed to reassure someone that a spot is safe.

DERM: AI triage inside the NHS

DERM, made by the UK company Skin Analytics, analyses close-up dermoscope photos of a lesion taken with a smartphone and a lens attachment, usually at a community clinic after a GP has made an urgent referral. If DERM classifies the lesion as benign, the patient can be discharged from the urgent pathway with advice to keep watching it. If it suggests a pre-cancer or cancer, an NHS dermatologist reviews the case.

DERM is the first dermatology AI to hold a Class III CE mark under the EU Medical Device Regulation, which allows autonomous use. In May 2025 NICE published an early value assessment (HTE24) that conditionally recommends DERM for NHS teledermatology for a three-year evidence-generation period. NHS England says it could discharge around 30% of cases from the urgent pathway.

The numbers behind the recommendation

Two details matter for patients. First, when clinicians did a second read of cases DERM had marked for discharge, they overturned 40% to 50% of those decisions; NICE's external assessors estimated a second read could lift sensitivity from about 95% to about 98%. Second, around a quarter of lesions are not eligible for DERM (for example because of their location or image quality) and need the usual route.

Skin tone safeguard

The two main prospective studies had almost no participants with black skin (0% and 1%). The company reported no missed cancers in Fitzpatrick skin types V and VI, but NICE judged the data too small. Its condition is clear: a healthcare professional must review DERM assessments for people with black or brown skin, and NHS England asks sites to start with a second read for everyone and to obtain informed consent before images are analysed.

What it costs the NHS

NICE's assessment used a price of £30 per referral for the DERM platform, plus £17 if the company's dermatologist provides the second read. For patients in the NHS there is no charge. Skin Analytics says DERM is an investigational device in the United States.

Mole-checking apps: what the evidence says

Consumer apps let you photograph a spot and get a risk rating. The most rigorous independent review is a 2020 BMJ systematic review by Freeman and colleagues. It found nine studies of six apps. The best result for the leading CE-marked app, SkinVision, was 80% sensitivity and 78% specificity, but that came from one study that also used users' questionnaire answers; in other evaluations its sensitivity ranged from 26% to 88%. Some apps missed every melanoma in their test sets.

The authors concluded that current apps cannot be relied on to detect all melanomas or other skin cancers, that the studies were small and poorly designed (often using photos taken by clinicians, not by app users), and that real-world accuracy is probably worse. They also criticised the regulatory route at the time, because low-risk class apps could be CE-marked without independent review. App versions and regulations have changed since; check whether any app you use states a current medical device approval for your country, and treat a "low risk" result as no reason to skip a doctor's visit.

The skin-of-colour problem

Skin cancer is less common in darker skin but is often diagnosed later, and it can look different (for example, melanoma under the nails or on the soles). Most dermatology image collections used to train AI are dominated by lighter skin. A 2022 study in Science Advances by Stanford researchers built the Diverse Dermatology Images dataset of biopsy-confirmed cases across skin tones; leading dermatology AI models performed markedly worse on it than in their original reports and worse on darker skin, and the authors found that fixing this required diverse training data. This is why the FDA flags limited data for darker skin with DermaSensor, and why NICE requires human review for black and brown skin with DERM.

What to expect as a patient

• In the US: your primary care doctor may use a device such as DermaSensor on a spot that worries them. A positive result means referral to a dermatologist, not a diagnosis.
• In the NHS: after an urgent GP referral you may be asked to attend a photo appointment. You should be told if AI will analyse your images and asked for consent. If DERM finds the lesion benign you may be discharged with advice; you can ask whether a clinician reviewed the result.
• Elsewhere in Europe: availability depends on the country and the health service; ask your doctor or dermatology clinic whether an AI tool is used and how results are checked.
• At home: an app can remind you to monitor spots, but a reassuring score is not evidence that a spot is harmless.

Risks and realistic expectations

• False reassurance: the biggest risk with any "benign" AI result, especially from apps without independent validation.
• Over-referral: high-sensitivity tools like DermaSensor flag many harmless spots, which can mean anxiety, extra appointments and unnecessary biopsies.
• Unsuitable lesions: nails, palms, soles, genitals, scalp and scars are often excluded. These sites are exactly where some melanomas hide.
• Evidence still maturing: NICE's recommendation for DERM is conditional and will be reassessed after three years of data collection.

Questions to ask your doctor

• Is an AI tool being used on my skin lesion, and what is it approved for?
• Will a dermatologist or trained clinician check the AI result?
• Has this tool been tested on people with my skin type?
• If I am discharged, what changes in the spot should make me come back, and how quickly?
• Would a biopsy be the right next step instead?

Questions patients ask