Researchers in Sweden have developed artificial intelligence systems capable of identifying individuals at elevated risk of developing melanoma well before the disease manifests. This breakthrough relies solely on health information extracted from existing medical records.
A recent study conducted at the University of Gothenburg reveals that AI can pinpoint people likely to develop this aggressive skin cancer within a five-year timeframe. The team analyzed clinical registry data encompassing nearly six million Swedish adults residing in the country from 2005 to 2014.
By incorporating extensive clinical details beyond basic age and gender—such as medication histories and previous diagnoses—the AI models demonstrated significantly enhanced predictive accuracy. The most sophisticated model correctly identified future melanoma cases with about 73 percent accuracy, outperforming the 64 percent accuracy achieved when only age and gender were considered.
Through integrating diverse sociodemographic and medical variables, the researchers isolated small subpopulations facing a 33 percent chance of developing melanoma within five years. Doctoral candidate Martin Gillstedt noted that although this decision-support tool is not yet part of standard healthcare practice, the findings highlight the strategic value of registry data in risk prediction.
Melanoma, mainly triggered by ultraviolet exposure from sunlight or tanning beds, represented four percent of all new cancer cases in the European Union in 2020. Lead author Sam Polesie emphasized that targeted screening of these identified high-risk groups could improve monitoring precision and optimize healthcare resource allocation.
Such an approach would enable clinicians to concentrate follow-up efforts and invite individuals deemed at elevated risk for screening via digital communication methods. While AI models trained on large-scale registry data show promise for personalized risk assessment, the researchers caution that further policy considerations are required before these tools can be implemented in routine clinical settings.
