Research & Evidence

ScanSkinAI Research, Clinical Evidence and Technical Publications

ScanSkinAI publishes clinical validation, technical methodology, product-transparency information and sourced public-health data to support responsible understanding of AI-assisted skin screening.

These resources explain how specific ScanSkinAI modules were developed and evaluated, what the reported results mean, and the limitations that must be considered.

View the Zenodo Publication

ScanSkinAI is intended to support preliminary skin awareness and screening. It does not provide a medical diagnosis and does not replace a doctor or dermatologist.

Quick answer

ScanSkinAI research is published in four categories: clinical validation (the Accuracy Report, an independent physician case-by-case review), technical methodology (how the model was trained and evaluated), product transparency (what the Cancer Flag Module outputs and where human oversight is required) and sourced public-health data. Reports are citable, version-dated and archived on Zenodo with a permanent DOI. ScanSkinAI provides preliminary screening information only — it is not a medical diagnosis.

Publisher
Ivy AI Solutions Limited (ScanSkinAI)
Evidence categories
Clinical validation, methodology, product transparency, public-health data
Permanent archive
Zenodo DOI 10.5281/zenodo.21225287
Last reviewed
2026-07-29
Next review
2027-01-29
Access
Free to read and cite; not a medical diagnosis

Four categories of evidence

Each category answers a different question. They are not interchangeable, and results from one should not be read as results from another.

  • Clinical Validation

    How ScanSkinAI outputs performed when reviewed case by case against clinical assessments by an independent physician reviewer.

    ScanSkinAI Accuracy Report
  • Technical Methodology

    How the model was designed, trained, evaluated and documented — data preparation, architecture, evaluation approach and explainability.

    AI Skin Cancer Screening Methodology
  • Product Transparency

    What one specific module does, what it outputs, where human oversight is required and where its intended-use boundaries end.

    Cancer Flag Module
  • Public-Health Data

    Sourced incidence, mortality, survival and trend data from recognised national cancer registries and public-health authorities.

    Skin Cancer Statistics

Clinical Validation

Clinical Validation Report

ScanSkinAI Accuracy Report

Review the independent clinical audit of ScanSkinAI screening outputs, including the study design, dataset, reviewer methodology, reported findings, triage performance and limitations.

Evaluation type
Independent dermatologist audit (Phase 1), retrospective review of production scans
Number of reviewed cases
100 anonymised cases
Conditions represented
80 skin conditions
Skin-tone representation
Fitzpatrick I–VI: I 7, II 14, III 21, IV 20, V 13, VI 5 cases (weighted toward III–IV)
Reviewer
Dr. Anand S. Urhekar, MD — independent physician reviewer
Publication date
6 July 2026
Last review date
6 July 2026
Headline result (as reported)
95.3% clinically acceptable accuracy — a screening metric, not diagnostic accuracy
Triage performance
8 over-triage and 2 under-triage cases out of 100 (2% under-triage rate)
Main limitations
Single-reviewer Phase 1 audit of 100 cases; photograph-based; not a diagnostic study; results apply only to the model version and dataset described in the report

View the Evidence Overview

Technical Methodology

Technical Whitepaper

AI Skin Cancer Screening Methodology

Explore the technical methodology used to develop and evaluate ScanSkinAI’s AI-assisted skin-screening technology, including data preparation, model architecture, evaluation approach, explainability and limitations.

Resource type
Company-authored technical whitepaper, archived on Zenodo with a permanent DOI
Author / publisher
Ivy AI Solutions Limited (developer of ScanSkinAI)
Publication year
2026
Model architecture described
DINOv2-Large vision transformer (ViT-L/14) at 518×518 resolution
Explainability
Grad-CAM and SHAP attention inspection
Quality framework
ISO 13485 quality management and ISO 27001 information security

Citation

Ivy AI Solutions Limited. (2026). ScanSkinAI Cancer Flag Module: Training Methodology, Model Development, and Robustness Evaluation. Zenodo. https://doi.org/10.5281/zenodo.21225287

Resource type: Archived technical whitepaper (Zenodo, permanent DOI). This is a company-authored technical report archived with a permanent DOI. It has not undergone independent academic peer review.

Featured Report

Digital Skin Health Insights Report 2026

A descriptive analysis of 50,265 historical scan-level risk-band records and 45,270 medical skin-module AI-output records, with clear interpretation limits and methodology.

Product and Model Transparency

Product Transparency Report

Cancer Flag Module

Understand the intended use, model workflow, supported output categories, image-quality safeguards, explainability, responsible communication and limitations of the ScanSkinAI Cancer Flag Module.

Intended use
Non-diagnostic screening and triage support from an ordinary clinical photograph
Module scope
One module within ScanSkinAI — not the wider multi-condition screening system
Output type
Four categories: basal cell carcinoma, melanoma, squamous cell carcinoma, and benign or other lesions, communicated as a triage signal
Image-quality requirements
Image-quality safeguards gate low-quality photographs before a flag is returned
Human oversight
Clinician review is required; the module supports prioritisation, it does not decide care
Non-diagnostic status
Does not replace dermoscopy, biopsy, histopathology or in-person examination
Important limitations
Photograph-based; evaluated on the datasets and model version described in the whitepaper; not validated for every population, lesion type or imaging condition

Modules are evaluated separately

  • Different modules may use different architectures.
  • Different modules may cover different output categories.
  • Different evaluations may use different datasets and metrics.
  • Results from one module must not automatically be applied to another module.

Public-Health Data

Public-Health Data Reference

Skin Cancer Statistics

Explore sourced skin-cancer incidence, mortality, survival and trend data from recognised national cancer registries and public-health authorities.

Countries covered
United Kingdom, United States and Australia
Types of statistics included
New cases, deaths, five-year survival, lifetime risk and UV-attributable share
Main data sources
Cancer Research UK; American Cancer Society and SEER; AIHW and Cancer Council Australia
Last reviewed
July 2026
Reference methodology
Every figure on the page carries an inline citation to its originating registry or authority

Public-health statistics describe population-level patterns. They do not represent ScanSkinAI users, ScanSkinAI screening results or individual medical risk.

How to Interpret ScanSkinAI Evidence

Different evidence sources answer different questions.

Clinical validation asks
How did the system perform when reviewed against a defined set of clinical cases?
Technical methodology asks
How was a model designed, trained, evaluated and documented?
Product transparency asks
What does a specific module do, and what are its intended-use boundaries?
Public-health data asks
What do recognised registries report about disease incidence, mortality and survival?

Results from one evaluation should not be generalised beyond the model version, dataset, population, output categories and study design described in the relevant report.

Research Governance

Research and content owner
Ivy AI Solutions Limited (ScanSkinAI)
Medical reviewer (this hub page)
Content governance review pending
Technical reviewer (this hub page)
Content governance review pending
Original publication date
29 July 2026
Last reviewed
29 July 2026
Next scheduled review
29 January 2027
Version
Hub page v1.0
Correction contact
info@scanskinai.com

This hub page summarises and links to the underlying reports. Reviewer information for each individual report is stated on that report. The Accuracy Report was reviewed independently by Dr. Anand S. Urhekar, MD; that review applies to the audit described in that report only and does not constitute review of this hub page. Citation guidance is provided in the Citing ScanSkinAI Research section below.

Research limitations: the evaluations linked here are limited in sample size, population coverage and model version. None of them establish diagnostic performance.

Transparency and Commercial Interest

  • ScanSkinAI, developed by Ivy AI Solutions Limited, builds and commercialises the technology described in these resources.
  • Some of these reports are company-authored, including the technical whitepaper and product-transparency documentation.
  • Where an evaluation was carried out by an independent reviewer, that is stated explicitly in the relevant report.
  • Company-authored methodology documentation is not the same as independent validation.
  • Commercial relationships do not change the non-diagnostic intended use of the product.
  • Read the full methodology and limitations in each report before interpreting any reported result.

Citing ScanSkinAI Research

  • ScanSkinAI Cancer Flag Module: Training Methodology, Model Development, and Robustness Evaluation

    Ivy AI Solutions Limited · 2026 · Archived technical whitepaper (Zenodo, permanent DOI)

    10.5281/zenodo.21225287

    Ivy AI Solutions Limited. (2026). ScanSkinAI Cancer Flag Module: Training Methodology, Model Development, and Robustness Evaluation. Zenodo. https://doi.org/10.5281/zenodo.21225287

  • ScanSkinAI Accuracy Report 2026

    Ivy AI Solutions Limited · 2026 · Clinical validation report (independent dermatologist audit)

    scanskinai.com/accuracy-report

    ScanSkinAI Accuracy Report 2026, Ivy AI Solutions Limited. https://www.scanskinai.com/accuracy-report

  • Skin Cancer Statistics 2026: UK, US & Australia

    ScanSkinAI · 2026 · Public-health data reference

    scanskinai.com/skin-cancer-statistics

    ScanSkinAI (2026). Skin Cancer Statistics 2026: UK, US & Australia. https://www.scanskinai.com/skin-cancer-statistics (Accessed [Month] [Year]).

Future Research

ScanSkinAI may publish additional responsibly aggregated digital skin-health insights after completing the necessary methodology, privacy, governance and clinical-review processes.

Research & Evidence: Frequently Asked Questions

ScanSkinAI publishes a clinical validation report (the Accuracy Report), an AI skin cancer screening methodology whitepaper, a Cancer Flag Module product-transparency report, and registry-sourced skin cancer statistics. Each is version-dated and independently citable.

ScanSkinAI outputs have been reviewed case by case against clinical assessments by an independent physician reviewer, and the findings are published in the Accuracy Report. This is company-published validation, not peer-reviewed academic research, and results apply only to the model version, dataset and population described in the report.

Cite the archived Zenodo whitepaper as: Ivy AI Solutions Limited. (2026). ScanSkinAI Cancer Flag Module: Training Methodology, Model Development, and Robustness Evaluation. Zenodo. https://doi.org/10.5281/zenodo.21225287. Copy-ready citations for each report are listed in the citations section of this page.

Yes. The Cancer Flag Module technical whitepaper is permanently archived on Zenodo under DOI 10.5281/zenodo.21225287, and the Digital Skin Health Insights Report 2026 has its own Zenodo DOI.

No. ScanSkinAI provides preliminary screening and skin-awareness information only. It does not diagnose disease, cannot rule out disease, and does not replace assessment by a doctor or dermatologist. Any new, changing, bleeding or non-healing skin mark should be reviewed by a qualified healthcare professional.

Research is produced by Ivy AI Solutions Limited, the developer of ScanSkinAI, with clinical input from qualified medical reviewers. Publications state their version, review date and next scheduled review date.

Medical and Research Disclaimer

The information presented in the ScanSkinAI Research & Evidence library is provided for research transparency and general educational purposes.

ScanSkinAI is designed to support preliminary skin awareness and screening. It does not provide a medical diagnosis, rule out disease or replace assessment by a doctor or dermatologist.

Performance results apply only to the model version, dataset, evaluation method and population described in each report. Results from one module or evaluation must not be generalised to all ScanSkinAI products or all users.

Company-authored technical documentation is not equivalent to independent clinical validation or peer-reviewed academic research.