• Faster RCNN
• DETR
• YOLO
• Oral squamous cell carcinoma
• OSCC
• Object Detection

The deep learning-based techniques designed in recent years are achieving the highest results in performing object detection task and are increasingly introduced in the computer-assisted image analysis systems. Oral squamous cell carcinoma (OSCC) is a malignant disease with a high rate of tumor metastasis; its early and accurate identification can enhance patient life and outcomes. In this work, a comparison between different deep learning based object detector to perform the lesion detection tasks is proposed in the context of oral cancer. Due to the many privacy issues arising when dealing with medical images, a novel image dataset has been collected and annotated in collaboration with medical centers. The main contribution of this work is to provide an overview of the the state-of-the-art of object detection architecture and present a performance comparison of such models on the self-built oral cancer dataset. To identify oral lesions via bounding boxes, a fine-tuning approach has been made to the state-of-the-art architectures: Faster R-CNN, YOLO, and DETR. The final goal is to identify the best deep architecture on which to design computer-assisted systems for oral squamous cell carcinoma screening and early cancer detection.