Text this: Comparison between observer-based and AI-based reading of CBCT datasets: An interrater-reliability study