Systematic Review: AI for Third Molar Imaging Analysis

A systematic review of 32 studies finds AI, mostly convolutional neural networks on panoramic X-rays, shows promising but variable performance for third molar detection, segmentation, classification, and surgical difficulty prediction; external validation is lacking.

• Of 227 screened records, 32 studies met criteria; most used CNN/YOLO architectures and panoramic radiography.
• Performance varied, with greater variability for complex tasks like root morphology, angular measurements, and surgical difficulty prediction.
• Methodological heterogeneity, inconsistent metrics, and limited external validation limited evidence.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.