AN INTELLIGENT DENTAL CYST PREDICTION FRAMEWORK USING PYRAMID SWIN VISION TRANSFORMER WITH REINFORCEMENT LEARNING - DRIVEN DIAGNOSTIC OPTIMIZATION

Abstract

Radiographic imaging is essential for the early and accurate prediction of dental
cysts to prevent jaw destruction, tooth displacement, and delayed clinical treatment. Existing
diagnostic methods rely heavily on expert interpretation and handcrafted features, resulting in
inter-observer variability and poor generalization across different imaging conditions.
Although recent Deep Learning (DL) approaches improve detection performance, they often
fail to effectively capture multi-scale pathological patterns and lack adaptive optimization of
diagnostic decisions. These limitations highlight the need for an intelligent and automated
diagnostic framework. This paper proposes an intelligent dental cyst prediction framework
based on a Pyramid Swin Vision Transformer with Reinforcement Learning (PSViT-RL)
based diagnostic optimization.