• image processing
• detection classification
• case-based reasoning
• convolutional neural network
• computer vision
• OSCC
• oral ulcer
• oral oncology
• diagnostic cancer

Oral Squamous Cell Carcinomas (OSCC) accounts for 90% of all oral cancers, it is the sixth most common cancer worldwide, and it is characterized by significant mortality and morbidity, for its late diagnosis and the impact of therapies on the patient’s quality of life. In the early detection of oral cancer, dental professionals play an important role by doing periodic thorough oral examinations and by using embedded smart cameras for oral photo and remote diagnosis. Deep Learning (DL) shows a higher potential for automated detection and classification of oral lesions. Case-based reasoning (CBR) is an artificial intelligence (AI) approach that focus on recognizing the similarity of past experience to solve current problems. DL and CBR can be used as a clinical tool for fast screening, earlier detection, and therapeutic efficacy assessment of the cancer. This thesis proposes an effective method, based on the combination of DL and CBR, to allow the post-hoc explanation of the system answer. The study illustrates the motivation for this approach, the design of the decision process, and the development of the DL-CBR decision support system. The DL engine is based on Feature Pyramid Network Resnet- 50 Faster R-CNN for object detection and classification, pre-trained with the Microsoft Common Objects in Context (COCO) data. CBR is based on 160 cases belonging to three classes of oral ulcers: neoplastic, aphthous and traumatic. The DL achieves the state-of-the-art performance, i.e., 82% detection and 90% classification rate (98% for neoplastic vs. no neoplastic binary classification), whereas the DL-CBR decision process is able to correctly address even less experienced doctors when dealing with borderline cases.