A Comparative Study of Deep Learning Models for Dental Segmentation in Panoramic Radiograph

Introduction: Dental segmentation in panoramic radiograph has become very relevant dentistry, since it allows health professionals to carry out their assessments more clearly and helps them define the best possible treatment plan for patients. Objectives: In this work, a comparative study is carried with four algorithms (U-Net, DCU-Net, DoubleU-Net Nano-Net) that are prominent medical literature on we evaluate results current state of art dental radiograph. Methods: These were tested dataset consisting 1500 images, considering experiment scenarios without augmentation data. Results: was model presented among analyzed models, reaching 96.591% accuracy 92.886% Dice using data augmentation. Another stood Nano-Net augmentation; achieved close only 235 thousand trainable parameters, while (TSASNet) contains 78 million. Conclusions: The obtained work satisfactory present paths better effective process.