Feasibility study of AI for radiographic DMFT score from panoramic radiographs with external validation

An AI pipeline for automatic rDMFT scoring from panoramic radiographs achieved an ICC of 0.84 (95% CI 0.74-0.90) vs. two-dentist reference on external validation, with mean absolute error 2.38, though decayed tooth detection was unreliable.

• The pipeline was externally validated on 100 panoramic radiographs from an independent hospital.
• For missing teeth, exact count was correct in 42% of cases but over-called in 15 of 74 complete dentitions.
• Agreement for filled teeth was high (r=0.90), while decayed tooth detection was unreliable (r=0.34 vs. dentists' 0.16).

Automated summary

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