- 期刊
利用融合AI進行最佳化牙科全口X光片拍攝擺位之調控策略
Optimizing Dental Panoramic X-ray Positioning Strategies Using Fusion AI
摘要
本研究主要目的是利用融合人工智慧（fusion artificial intelligence, fusion AI）技術分析牙科全口X光片（PANO）擺位錯誤。透過轉移學習訓練後之CNN產生擺位錯誤之融合預測值，用以提供臨床全口牙擺位錯誤調整措施。因此，本研究將六種CNN模型經轉移學習後整合為融合AI系統，將其輸出之六種擺位錯誤影像機率值加總，得到錯誤種類的預測值，再執行相關性分析求得相關係數矩陣，用以評估錯誤擺位之樞紐機制，做為後續臨床全口牙擺位之調控策略與措施。實驗結果證明，融合AI系統對於不同擺位錯誤類別均具有良好且具統計顯著性的預測能力。此外，本研究提供擺位錯誤相關性路徑圖，通過分析錯誤之間的關聯性，找出關鍵樞紐點，提供放射師進行擺位改善或影像品管時提供校正策略。綜上所述，本研究成功地將多種AI分類模型融合，將能提高牙科全口X光片常見拍攝正確率，並為相關專業人士提供了有價值的參考資料和策略建議。
並列摘要
The main objective of this study is to use Fusion Artificial Intelligence (Fusion AI) technology to analyze misalignment errors in dental full-mouth X-rays (PANOs). A Fusion AI system is integrated by training six CNN models using transfer learning to generate fusion predictions of misalignment errors, which can be used to provide clinical adjustment measures for full-mouth dental misalignment. The probabilities of the six misalignment error images output by the system are summed to obtain the predicted values of the error types. Relevant correlation analysis is then performed to obtain the correlation coefficient matrix, which is used to evaluate the pivotal mechanism of error misalignment and to develop subsequent strategies and measures for clinical full-mouth dental alignment. The experimental results show that the Fusion AI system has good and statistically significant predictive capabilities for different misalignment error categories. In addition, the study provides a misalignment error correlation path diagram to identify key pivotal points and provide correction strategies for radiologists to improve the accuracy of image quality control during dental misalignment improvement. In summary, this study successfully integrates multiple AI classification models to improve the accuracy of common dental full-mouth X-ray imaging, providing valuable reference data and strategy recommendations for relevant professionals.