A major study published in Acta Dermato-Venereologica has demonstrated that AI can identify individuals at increased risk of developing melanoma up to five years before diagnosis — using nothing more than routinely collected health registry data.
The Study
Conducted by researchers at the University of Gothenburg in collaboration with Chalmers University of Technology, the study analysed national registry data from over 6 million Swedish adults.
The AI models processed:
- Age and gender
- Prior medical diagnoses
- Medication history
- Socioeconomic factors
By identifying subtle patterns across these data points, the models surfaced risk signals that conventional screening methods miss entirely.
The Results
- The best-performing AI model correctly distinguished between individuals who would later develop melanoma and those who would not in approximately 73% of cases
- For comparison, using only age and gender achieved about 64% accuracy
- The models identified smaller, high-risk subgroups where the probability of developing melanoma within five years reached approximately 33%
Clinical Potential
The most significant aspect of this research is what it doesn’t require. There are no genetic tests, no imaging, no specialist appointments. The data already exists in healthcare systems worldwide. The AI simply identifies risk patterns that are invisible to human analysis at population scale.
If validated in clinical practice, this approach could:
- Focus screening resources on the highest-risk individuals rather than broad population sweeps
- Enable earlier detection by flagging patients for monitoring years before symptoms appear
- Reduce healthcare costs by replacing expensive universal screening with targeted surveillance
Caveats
The researchers are clear that this is not yet ready for clinical deployment. Further validation, policy decisions, and additional research are needed before AI-based melanoma risk assessment can be integrated into standard healthcare workflows.
But the signal is strong: routine data, properly analysed, contains predictive power that current medical practice leaves on the table.
Source: sciencedaily.com, gu.se, eurekalert.org
