• case-based reasoning
• deep learning
• explainable artificial intelligence
• oncology medical imaging
• oral cancer
• self-supervised learning
• supervised learning

Recognition of oral squamous cell carcinoma (OSCC) is a particularly challenging task due to the often late diagnosis of the disease. In this context, a screening system based on deep learning (DL) models presents a promising solution for the early detection of OSCC. DL models have shown remarkable success in various medical imaging applications due to their ability to learn complex patterns and features from large datasets. However, the lack of transparency and interpretability of these models significantly hinders their widespread adoption in medical practice.
Explicable artificial intelligence (XAI) provides techniques for understanding models. However, current XAI is mostly data-driven and focuses on developers’ needs to improve models rather than users’ requests to express relevant insights.
Among different XAI strategies, we explore a solution composed of the Case-Based Reasoning paradigm to provide visual explanations of results and DL to handle photographic images, resulting in a DL-CBR system. We conducted several experimental benchmarks on an oral cavity dataset collected in collaboration with medical centers. Our DL-CBR approach achieves performance comparable to pure DL models but without the issue of inexplicability, making it more suitable for practical clinical use.