Beyond Ground Truth: A Heuristic Unsupervised Quality Assurance Engine for Diagnosing Calibrational Blindness in Dental Segmentation

This research proposes a novel heuristic unsupervised quality assurance engine designed to detect and diagnose calibrational blindness in dental segmentation models without relying strictly on ground truth. By addressing the limitations of traditional verification methods, the study introduces an automated evaluation framework developed specifically for dental radiography. It also shows how the framework improves the reliability, transparency, and practical applicability of AI-driven dental diagnostic support tools for real-world clinical use.