ADVANCEMENTS IN AI-DRIVEN SEGMENTATION AND CLASSIFICATIONS FOR ORAL EPITHELIAL DYSPLASIA IMAGES

This review examines the transformative potential of artificial intelligence (AI), machine learning (ML), and deep learning (DL) in improving the diagnosis of oral epithelial dysplasia (OED), a global issue and a critical concern in India, where over 48% of the population is affected by oral cancers. The study highlights the limitations of manual diagnosis and variability in assessment criteria, focusing on AI and DL's role in histological grading (HG) and segmentation of OED. These technologies can detect subtle patterns in histopathological images that pathologists may miss. AI-enabled OED detection primarily involves two key steps: segmentation and classification of regions-of-interest (RoI) in histopathological images. Segmentation allows AI algorithms to focus on abnormal tissue areas, improving diagnostic accuracy. DL tools then classify OEDs by identifying complex patterns, enhancing grading consistency. From an analysis of 236 papers, 98 were relevant, with 17 domain-specific studies on AI-driven segmentation and classification of OED selected. The systematic time-lined review explores various AI and DL methodologies, comparing their strengths and limitations, while emphasizing the need for collaboration among researchers, dentists, and clinicians to enhance early detection and treatment of head and neck cancers. These advancements hold great promise for improving survival rates and patient quality of life.