Methodology

How Skinvest Works

The AI behind our triage tool, our limitations, and our clinical oversight.

Medically reviewed by our clinical advisory group of practising dermatologists · Last reviewed: 5 May 2026

Intended Use

Skinvest is designed as an AI-powered skin triage tool. It analyses a user-submitted image and returns insights about likely skin presentations, intended to help users decide whether a concern warrants evaluation by a qualified dermatologist. Skinvest is not a diagnostic platform and does not return a clinical diagnosis. Skinvest is a SAHPRA-licensed Class B Software as a Medical Device (SaMD).

What Skinvest Is — and Isn't

Skinvest is:

  • A triage tool — output is an indicative starting point, not a verdict
  • An awareness tool — helps users learn what to watch for and when to seek care
  • A SAHPRA-regulated medical device, operated under Class B SaMD standards

Skinvest is not:

  • A diagnostic platform — it does not produce a medical diagnosis
  • A substitute for a dermatologist consultation, biopsy, or any clinical examination
  • A "skin cancer detector" — Skinvest covers a broad set of skin presentations, not a cancer-specific screening tool

Model Approach

Skinvest uses deep convolutional and modern vision architectures to classify uploaded skin images into a set of trained presentations. The training pipeline draws on dermatological imagery from public, licensed, and partner datasets, processed under our regulatory data-handling controls. Outputs include the most-likely class together with an internal confidence indicator. We deliberately avoid presenting confidence as a clinical probability of disease — see "Limitations" below.

Free Web Demo vs Full SaMD Product

The free web demo at skinvest.ai/run_inference uses a constrained model (~50 trained classes, no "inconclusive" output). It is intended to demonstrate the technology and to support skin-health awareness. The full Skinvest SaMD product, operated under our SAHPRA licence, uses materially larger datasets, expanded condition coverage, and includes "inconclusive" output handling where appropriate.

Known Limitations

  • Forced-classification behaviour (free demo): The demo will always return its closest training-set match, even when the input image is not skin or shows a condition outside its training scope.
  • Confidence is not a clinical probability: The displayed confidence reflects how closely the input resembles the nearest training example; it is not a probability of disease.
  • Image quality matters: Lighting, focus, angle, occlusion (hair, jewellery), and image compression affect output. Bad inputs produce unreliable outputs.
  • Skin-tone coverage is an active commitment, not a finished claim: Public dermatology datasets are well-documented to over-represent lighter skin tones. We are actively working to validate and improve performance across the full Fitzpatrick I–VI range, with input from our medical advisors.
  • Out-of-distribution inputs: Images of unusual presentations or conditions outside the training set may produce confidently-wrong outputs in the free demo.

Clinical Oversight

Skinvest's content, triage approach, and educational material are reviewed by an active clinical advisory group of practising dermatologists. Advisors consult on triage refinement, content review, and inclusivity considerations across the Fitzpatrick I–VI spectrum. Out of respect for our advisors' privacy, we do not publish their names. Clinical credentials, scope of work, and correspondence with regulators are available on request to qualified parties.

Regulatory Posture

Skinvest is licensed by the South African Health Products Regulatory Authority (SAHPRA) as a Class B Software as a Medical Device (SaMD). SaMD operations are subject to ongoing post-market surveillance, change-control procedures, and incident reporting under our regulatory obligations.

What to Do With a Skinvest Result

  1. Treat the output as an indicative starting point, not a diagnosis.
  2. Use it to help frame a conversation with a qualified dermatologist or healthcare provider.
  3. Seek prompt professional evaluation for any lesion that is changing, bleeding, painful, or causing concern — regardless of what Skinvest says.
  4. Use the educational content (e.g., the ABCDE rule on the home page) to support — not replace — that conversation.

Reporting Issues

If you believe Skinvest produced a materially misleading output, or you experienced an issue with the service, please email support@skinvest.ai. As a SAHPRA-regulated operator, we take post-market feedback seriously and feed it into our incident- and change-control processes.

Methodology page reviewed by Skinvest's medical advisory board. Last reviewed: 5 May 2026.