AI Can Predict Melanoma Risk Up to Five Years Before Diagnosis

Predicting cancer before a single physical symptom appears has long been a goal of preventative medicine. Now, researchers have made a significant leap forward. A study from the University of Gothenburg has demonstrated that artificial intelligence (AI) can identify risk patterns for melanoma in adults up to five years before the cancer develops.

Unlike traditional screening methods that rely on visual inspections of moles or lesions, this new approach uses massive datasets to spot subtle markers of vulnerability. By shifting the focus from reactive treatment to proactive prevention, this technology could fundamentally change how clinicians monitor patients and allocate healthcare resources.

Moving Beyond Images: The Power of Registry Data

Most AI tools used in dermatology today are image-based, meaning they analyze photos of skin lesions to determine if they are malignant. While useful, this happens after a physical change has already occurred.

The University of Gothenburg study took a different path by using healthcare registry data. The AI analyzed information from 6 million adults in Sweden, looking at population-level data to uncover correlations that precede the physical manifestation of the disease. The models examined several key variables, including:

• Age and sex
• Previous medical diagnoses
• Medication use
• Socioeconomic status

By processing these variables, the AI can isolate specific patterns that signal an increased probability of developing melanoma, allowing health systems to flag high-risk individuals long before a lesion appears.

The Results: Accuracy and Early Warning

The study tracked 6 million adults over a five-year period, during which 38,582 people (0.64%) developed melanoma. The researchers compared a basic predictive model against their advanced AI model to test its effectiveness.

Comparing Predictive Models

A basic model using only age and sex correctly identified melanoma cases 64% of the time. However, the advanced AI model—which incorporated diagnoses, medications, and sociodemographic data—reached 73% accuracy.

Identifying High-Risk Subgroups

The most critical finding was the AI’s ability to isolate small, high-risk population subgroups. For these specific groups, the probability of developing melanoma within a five-year window jumped to approximately 33%. This five-year prediction window provides a vital opportunity for early intervention and intensified monitoring.

The Shift Toward Precision Dermatology

This breakthrough represents a move toward precision medicine. Instead of screening the entire population with the same frequency, healthcare providers could use these risk assessments to target their efforts more efficiently.

Sam Polesie, Associate Professor of Dermatology and Venereology at the University of Gothenburg, noted that selective screening of these small, high-risk groups could lead to more accurate monitoring and a more efficient use of healthcare resources. This approach combines population-level data with clinical judgment rather than replacing the doctor’s role.

Frequently Asked Questions

Does this AI replace skin checks with a doctor?

No. The research emphasizes that this is a tool for precision medicine that combines data with clinical judgment. It is designed to help clinicians decide who needs more frequent screenings, not to replace the visual exam.

How is this different from a genetic test?

While genetic tests look at DNA, this AI looks at “registry data”—the history of your health records, medications, and socioeconomic factors—to find patterns that correlate with future cancer risk.

When will this be available in clinics?

The researchers have emphasized that implementation still requires further research and policy decisions before it can be integrated into routine healthcare systems.

Looking Ahead

The ability to predict melanoma risk years in advance transforms the clinical approach from “detect and treat” to “predict and prevent.” As machine learning continues to integrate with existing health infrastructure, the goal is to ensure that the people most at risk receive the most attentive care, ultimately saving lives through earlier detection.