Background: Keratinocyte cancer (KC), primarily non-fatal, poses a significant public health challenge in Australia and is associated with high healthcare costs and morbidity. Artificial intelligence (AI) advances offer promising tools for improving KC management.
Methods: This systematic review assesses recent AI applications in diagnosing and managing KC. We searched MEDLINE and EMBASE databases using terms like "squamous cell carcinoma," "basal cell carcinoma," "keratinocyte carcinoma," and "artificial intelligence," focusing on studies involving human participants published in English over the last five years.
Results: Our search yielded 144 articles, with 21 meeting the inclusion criteria after excluding duplicates and irrelevant studies. These articles reveal significant AI advancements in various modalities and KC types: seven focus on BCC, five on SCC, and nine on combined KC types with other skin cancers. Notable developments include AI models like EfficientNET-B0 and SkinViT for Dermoscopy and confocal microscopy in BCC, intraoperative margin analysis for SCC, and hyperspectral microscopic imaging. AI has also played a crucial role in dermatopathology by identifying biomarkers and genetic indicators and developing educational tools to improve practitioner skills. Most of the studies have been on training models, not real-world examples, and need more rigorous testing.
Conclusion: AI technology holds great potential to transform KC management, providing innovative, accurate, and efficient diagnostic solutions. These advancements could reduce healthcare costs by minimizing invasive procedures and frequent consultations while enhancing medical education. Continuous evolution and integration of AI into clinical practice are expected to significantly improve patient outcomes in KC management.