|Abstract: ||假牙?復治療一直是牙科治療針對患者缺牙重建相當重要的一環，隨著時代變遷及科技進步，以鈦金屬牙根取代傳統式假牙製作的人工植牙手術更是成為全口全建的首選。但是因為植牙手術牽涉範圍相當廣泛，舉凡病人身體狀況、手術醫師個人經驗及植牙系統的選擇等，都是治療必須考量之項目。只是因為植牙治療金額相較於傳統式假牙高出許多，讓許多醫療人員趨之若鶩，但卻未仔細分析手術未來影響因素而導致醫療糾紛的發生。 目標：在過去文獻中探討植體成敗因素多數僅就患者系統性疾病、手術方式及?復型式等單一因素進行統計學上相關顯著分析;本研究將建立預警機制以減少失敗機會，且進一步納入醫師個別因素及?復體固定方式。 方法：本研究收集嘉義某區域教學醫院接受人工植牙治療患者臨床資料，共有患者年齡、性別、缺牙原因、系統性疾病、抽煙、喝酒、嚼檳榔、醫師科別、醫師執業年數、植牙位置、骨頭硬度、骨脊增生術、上顎竇增高術、植體系統、植體長度、植體寬度、?復物形式、支台體角度及義齒固定方式，總計八大類別20個變項，再以決策樹、支援向量機及邏輯斯迴歸等監督式學習技術分成植體早期成敗預測及晚期成敗預測兩個面向進行分析。研究結果：以單一分類器比較預測率方面，不論植體早期成敗或是晚期成敗方面，C4.5(J48)決策樹預測效能最高，分別為62.6%及69.8%；並從分析中發現植體系統不分植體早期成敗及晚期成敗皆有影響。另外又以CART分類迴歸樹建立早期及晚期風險預測模型，建立影響植體早期及晚期成功失敗的關聯規則。結論：本研究希望能藉由研究之結果建立植體早期及晚期成敗預測模型，幫助臨床醫師遇到臨床相關情形時，可以選用預測結果較佳之植體系統及?復治療，以減少治療失敗發生糾紛之目的。|
Prosthodontic treatment hasalways been a major part of dental treatment for rehabilitation of patient with edentulous ridge. The traditional dentures were replaced with dental implant surgery insertion due to scientific and technological progressing. Dental implant surgery involves range to be quite extensive, include physical condition of patient, personal habit, surgical experience , and the choice of implant system etc, so all of the projects that must be considered. Because charge of dental implant is better than other treatment , many dentists preferred to perform implant surgery without thinking about the complications which cause the emergence of the medical dispute. Purpose:In the past literature, most of the exploratory factors were statistically significant only for single disease such as systemic disease, surgical procedure and prosthetic type. We wants toestablish early warning mechanism including factors of dentist and fixation of denture to reduce the chance of failure.Methods:We collected clinical data of patients undergoing dental implant treatment in a local hospital of Chiayi, including age, sex, causes of tooth loss, systemic diseases, smoking, drinking, chewing betel nuts, department of dental division, physician practice years, implant position , Bone density, bony augmentation, maxillary sinus augmentation, implant systems, implant length, implant width, prosthetic form, the angle of the abutment and fixation of denture.A total of eight categories of 20 variables, Decision tree, support vector machine and logistic regression were used to analyze the early success or failure prediction and the late success or failure prediction. Results: C4.5 (J48) decision tree had the highest predictive efficiency of 62.6% and 69.8%respectively, regardless of the early or late successes and failures in a single classifier.In addition, the early and late risk prediction models were established based on the classification regression tree (CART) , and the association rules were established to affect the early and late success of the implant. Conclusion:Ｗe hope to establish the prediction models of early and late success or failure of the implant by using the results of the study, and help the clinicians to choose the optimal implant system and the prosthetic therapy to reduce failure of treatment.