A new artificial intelligence model has demonstrated performance on par with experienced dermatologists in assessing the aggressiveness of squamous cell carcinoma, a common and increasingly prevalent form of skin cancer. The study, led by researchers at the University of Gothenburg, highlights the potential for AI to support clinical decision-making in dermatology. Each year, over 10,000 people in Sweden are diagnosed with squamous cell carcinoma, the second most frequent skin cancer after basal cell carcinoma. It typically develops in sun-exposed areas such as the head and neck, often in skin that has already shown signs of long-term UV damage—like rough, scaly patches and uneven pigmentation. While diagnosing the cancer itself is usually straightforward, determining how aggressively it is growing is a complex challenge. This preoperative assessment is critical for planning surgery: more aggressive tumors require wider excision margins and prompt treatment, while less aggressive ones may allow for simpler procedures. In Sweden and many other countries, preoperative punch biopsies are not routinely used for suspected squamous cell carcinomas. Instead, surgery is performed based on clinical appearance, with the full tumor removed and later analyzed through histopathology. This makes non-invasive tools like AI-powered image analysis highly valuable. The research team trained an AI model using 1,829 clinical close-up images of confirmed squamous cell carcinomas collected from Sahlgrenska University Hospital between 2015 and 2023. The model was then tested on 300 new images and compared against assessments from seven experienced dermatologists. Results published in the Journal of the American Academy of Dermatology International showed that the AI model’s performance was nearly identical to that of the dermatologists. Notably, agreement among the human experts was only moderate, indicating the difficulty of the task even for specialists. The study also identified two key visual features linked to higher tumor aggressiveness: ulcerated skin surfaces and flat tumor topography. Tumors with these traits were more than twice as likely to be classified in the two most aggressive categories. Associate Professor Sam Polesie, who led the study and practices at Sahlgrenska University Hospital, stressed the importance of integrating AI into areas where it can genuinely improve patient care. While AI has drawn significant attention in dermatology, its real-world impact has so far been limited. “This model needs further refinement and validation, but our findings suggest a clear path forward: AI should be used where it adds measurable value to clinical decisions,” Polesie said. He emphasized that health care needs—not technological novelty—should guide AI implementation. The research underscores the potential of AI to enhance early, non-invasive assessment of skin cancer, potentially leading to faster, more accurate treatment planning and better outcomes for patients.