Research library
AI in dentistry
-
Endodontics & RestorativeAI Root Canal Healing: Can a Model Call It?
AI root canal healing predictions sound impressive — until a model scoring 80% accuracy caught 0% of failures. Here’s what the numbers really mean.
-
Ethics, Education & Patient CareAI Caries Prediction: Fair to Every Kid, or Biased?
AI caries prediction hit ~78% accuracy in 5,710 U.S. kids — but its error rates differed by race until a fairness fix narrowed the…
-
Diagnostics & ImagingMobile AI Oral Cancer Screening: Lifesaver or False Alarm?
Mobile AI oral cancer screening caught 75% of high-risk lesions in a real Taiwan field trial — but 5 of 6 red flags were…
-
Ethics, Education & Patient CareCan AI Virtual Patients Teach Dentists Empathy?
AI virtual patients out-taught peer role-play on dental students’ communication skills in a randomized trial — but can a machine build real empathy?
-
OrthodonticsAI 3D Face Scan: Can a Selfie Replace the Scanner?
AI 3D face scan from one photo matched a $30,000 scanner for precision in a 47-patient study — but a systematic downward chin bias…
-
Implants & Oral SurgeryPeri-Implantitis Prediction: Can AI Call It Early?
Peri-implantitis prediction from a baseline dental X-ray reached AUC 0.83 in a two-center AI study — promising, but it still missed 1 in 3…
-
Diagnostics & ImagingAI Sleep Apnea Screening: Can a Dental X-Ray Spot It in Kids?
AI sleep apnea screening scored 0.945 AUC reading pediatric OSA risk from a routine dental X-ray — a promising flag, not a diagnosis.
-
Diagnostics & ImagingAI Jaw Lesion Detection: A Safety Net With Gaps?
AI jaw lesion detection can flag cysts and tumours on a panoramic X-ray, but the best model still missed a quarter of lesions. Here’s…
-
Diagnostics & ImagingAI Maxillary Sinus Detection: 99% Accurate or Too Easy?
AI maxillary sinus detection hit 99.43% accuracy on CBCT scans — but on 874 images from one clinic, is near-perfect a breakthrough or a…
-
Generative AI & ChatbotsAI Denture Design: Reliable, or Different Every Time?
AI denture design by four chatbots: expert raters agreed, but each bot drafted a different partial every time the same case was asked.