歯科インプラント周囲の歯槽骨吸収のデンタルX線写真に基づく補助測定方法

To develop an efficient and accurate method for the precise quantification of peri-implant marginal bone loss, aiming to overcome the limitations of conventional assessments that rely heavily on expert subjective judgment and to enhance the monitoring of marginal bone level stability and facilitate the early detection of clinical risk factors.

This research proposes an efficient automatic quantification method for marginal bone loss based on the YOLOv8-pose model. By detecting and localising keypoints associated with peri-implant bone loss, this method enables automated identification of marginal bone loss characteristics. Model performance is evaluated using the marginal bone loss rate to assess its effectiveness in predicting the severity of resorption on both sides of the implant.

The precision of keypoint detection for marginal bone loss features exceeds 98%, and the mean radial error of six keypoints in most images is less than 10 pixels, with the minimum error reaching only 2.89 pixels. In predicting marginal bone loss severity on both implant sides, the model achieves accuracies of 0.906 and 0.844, respectively, demonstrating robust predictive capability and application potential.

The YOLOv8-pose-based method effectively addresses the inefficiency and subjec-tivity of traditional assessments by enhancing the automation and objectivity of peri-implant marginal bone loss detection. It provides valuable technical support for accurate evaluation and clinical decision making in peri-implant marginal bone loss.