Ensemble Approach: Multiple Models, One Result
DermCheck doesn't use just a single AI model, but an ensemble of multiple specialized models. Each model was trained on different datasets and brings different strengths. By combining all results, the analysis becomes more robust and reliable than any individual model alone.
The Three Classification Models
Three specialized computer vision models analyze your image in parallel. Each model outputs probabilities for different skin cancer types: basal cell carcinoma (BCC), melanoma, actinic keratosis, benign keratosis, dermatofibroma, vascular lesions, and nevus (common mole). Results are aggregated via majority voting and weighted averaging.
The Textual ABCDE Analysis
In addition to numerical classification, a vision-language model creates a detailed textual analysis following the ABCDE rule. It describes the asymmetry, border, color distribution, size, and possible changes in the skin lesion. This description makes the results understandable and traceable for you.
Consensus and Confidence
When all three classification models reach the same conclusion, confidence is high. If the models disagree, the consensus score drops and the recommendation tends toward medical clarification. This system ensures that the safe path is always recommended when there is uncertainty.
Data Privacy and Security
Your images are transmitted encrypted and processed in secure European data centers. You have the right to have your data deleted at any time. DermCheck complies with GDPR requirements and treats your health data as a specially protected data category under Art. 9 GDPR.
Limitations of the Technology
AI models are tools — not infallible diagnosticians. The quality of the analysis depends on image quality: good lighting, sharp focus, and the right distance are crucial. Certain skin changes (e.g., under nails or on the scalp) are difficult to photograph. When in doubt, you should always see a dermatologist.
