Feasibility of multi-class dental caries detection using deep learning–based smartphone images: a pilot prospective study

Description

Ji-Yun Kang and Hee-Kyung Kim test whether phone photos plus deep learning can grade tooth decay, in this 2026 pilot study. Seventy adults took six photos of their own mouths with smartphones, outside any clinic. Four dentists graded each photo under two standard systems, at two, four, and five levels of decay detail. A detection model was trained on expert-labeled images.

Dentists agreed well with themselves but only loosely with each other. The models reached high marks across schemes: sensitivity above eighty-six percent and accuracy above ninety-three. Finer grading lowered the detection score, while the five-level model had the best specificity. The authors present phone-based screening as a workable aid patients could drive themselves, pending larger trials.

Catalog record

Source: https://europepmc.org/articles/PMC13194414