A large-scale Swedish study shows that artificial intelligence (AI) can identify risk patterns for melanoma, the most aggressive form of skin cancer, up to five years before diagnosis. The researchers demonstrate that existing healthcare data, which is routinely collected, provides a valuable basis for early and personalized risk assessment.
The study, conducted by the University of Gothenburg in collaboration with Chalmers University of Technology, was published in the scientific journal Acta Dermato-Venereologica. The analysis utilized national registry data from over six million Swedish adults. Within this population, 0.64 percent (38,582 individuals) developed melanoma over a five-year period.
The dataset included age, gender, medical diagnoses, medication use, and socioeconomic factors. By combining this data with AI models, researchers were able to identify subtle patterns that had previously gone unnoticed. “Our study shows that data already available within healthcare systems can be used more strategically to identify individuals at increased risk,” said researcher Martin Gillstedt of the Sahlgrenska Academy. According to him, this type of data-driven decision-making is not yet standard practice, but the results offer clear starting points for future applications.
Higher predictive value
The study compared various AI models. The most advanced model proved capable of correctly predicting which individuals would later develop melanoma in approximately 73 percent of cases. By comparison, a model that relied solely on age and gender achieved an accuracy of about 64 percent.
By combining multiple data sources, the researchers were also able to identify small subgroups with a significantly increased risk. In these groups, the likelihood of developing melanoma within five years rose to approximately 33 percent.
According to study lead Sam Polesie, a dermatologist and associate professor, this opens up possibilities for targeted screening. “Selective screening of high-risk groups can lead to more accurate monitoring and a more efficient use of healthcare resources. This brings population data closer to the practice of precision medicine,” he states.
Implications for healthcare capacity
The findings underscore the potential of AI not only to support healthcare processes but also to proactively improve them. By identifying high-risk groups early on, healthcare providers can screen more effectively and potentially intervene sooner. This is particularly relevant given the rising incidence of skin cancer worldwide and the pressure on healthcare systems.
At the same time, the researchers emphasize that further validation and policy considerations are needed before implementation in clinical practice is possible. Aspects such as ethics, data privacy, and integration into existing workflows play a crucial role in this regard. The study illustrates how AI, fueled by large-scale healthcare data, can contribute to a shift from reactive to preventive care. Instead of treating patients exclusively after diagnosis, there is now room for early intervention based on individual risk profiles.
According to the researchers, such models could play an important role in the future in personalized screening strategies for melanoma and other conditions. This research thus marks the next step in the development of data-driven, predictive healthcare.
Skin cancer diagnosis innovation
Last year an international study led by Karolinska Institutet in collaboration with Yale University shows that AI can help pathologists assess skin cancer tissue samples more consistently and accurately. The tool focuses on analysing tumour-infiltrating lymphocytes (TILs), key biomarkers that indicate how aggressively melanoma is developing. By objectively quantifying TILs from digital images, AI reduces subjective interpretation, a known limitation in pathology.
In the study, 98 participants evaluated melanoma tissue samples, with one group using standard methods and another supported by AI. Results showed improved reproducibility and better prognostic accuracy with AI assistance. Although relatively small and retrospective, the findings highlight AI’s potential to enhance diagnostic reliability and support more precise treatment decisions in cancer care.
