• deep learning
• oral cancer
• semantic segmentation

Medical image segmentation is an important task in assisted diagnosis and screening systems in several medical areas including oral cancer recognition. In this paper, we explore the effectiveness of different deep learning (DL) architectures, including FCN, Deeplab and Unet, for medical image segmentation related to oral cancer. Furthermore, we propose an ensemble model that incorporates several decision fusion strategies to aggregate individual predictions, to improve the individual model performance.
Our study employs a dataset acquired and manually labeled by the clinical subgroup of our team. On this dataset, we address two distinct segmentation problems: binary semantic segmentation to differentiate healthy tissue from diseased regions and multiclass semantic segmentation to identify three specific oral pathologies: aphthous, traumatic, and neoplastic lesions.
Furthermore, we study the ensemble model's effectiveness in improving segmentation accuracy by combining different DL architectures' strengths. The results obtained demonstrate an ensemble strategy is effective for the binary semantic segmentation problem, but does not improve the multiclass problem where we distinguish between different pathologies.