从口内X线片识别牙种植体的AI系统：诊断准确性研究

STATEMENT OF PROBLEM: Identifying an implant system from radiographic images is essential for managing prosthetic complications but often relies on a clinician's subjective assessment and incomplete patient records. The increasing number of implant systems worldwide complicates accurate recognition and delays treatment procedures.

PURPOSE: This study aimed to evaluate the diagnostic performance of a commercially available artificial intelligence software program in identifying different dental implant models from periapical radiographs.

MATERIAL AND METHODS: An observational retrospective study was conducted using 720 anonymized periapical radiographs representing 12 implant systems. Each radiograph was uploaded into the MovumStudio implant identification module, and a response was considered correct when the software program identified the implant brand or model as the first option with ≥70% confidence. Descriptive and inferential statistics were performed to assess accuracy, sensitivity, specificity, predictive values, the Cohen kappa coefficient, and the area under the receiver operating characteristic curves (α=.05).

RESULTS: The artificial intelligence system correctly identified 589 of 674 valid intraoral radiographs, achieving an overall accuracy of 87.4% (95% confidence interval, 84.9% to 89.9%) and a Cohen kappa value of 0.864, indicating almost perfect agreement. Sensitivity and specificity were ≥90% for most implant systems, with the area under the receiver operating characteristic curve above 0.90, except for the GMI Frontier system (area under the curve=0.55). The mean response time was 8.42 ±5.96 seconds.

CONCLUSIONS: The MovumStudio implant identification module demonstrated high diagnostic accuracy and reliability for recognizing most implant systems from periapical radiographs, offering a rapid, user-independent tool to support clinical decision-making.